Breast Cancer Research and Treatment 68: 45–54, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.

Report

The evaluation of human breast lesions with magnetic resonance imaging and proton magnetic resonance spectroscopy Kim M. Cecil1,2 , Mitchell D. Schnall1 , Evan S. Siegelman1, and Robert E. Lenkinski3 1 Department

of Radiology, University of Pennsylvania Medical Center, Philadelphia, PA; 2 Present address: Department of Radiology, Children’s Hospital Medical Center, Imaging Research Center, Cincinnati, OH; 3 Department of Radiology, Harvard Medical School, Beth Israel Deaconess Medical Center, Boston, MA, USA

Key words: breast cancer, choline, human, magnetic resonance imaging, neoplasm, spectroscopy

Summary Purpose. MR spectroscopy (MRS) assists in lesion characterization and diagnosis when combined with magnetic resonance imaging (MRI). Cancerous lesions demonstrate elevated composite choline levels arising from increased cellular proliferation. Our study investigated if MR spectroscopy of the breast would be useful for characterizing benign and malignant lesions. Materials and methods. Single voxel proton MR spectroscopy (MRS) was acquired as part of an MR imaging protocol in 38 patients referred upon surgical consultation. The MR spectra were read independently in a blinded fashion without the MR images by three spectroscopists. The MRI exam was interpreted in two settings: (a) as a clinical exam with detailed histories and results from previous imaging studies such as mammography or ultrasound included and (b) as a blinded study without prior histories or imaging results. Results. Elevated choline levels were demonstrated by MRS in 19 of the 23 confirmed cancer patients. The sensitivity and specificity for determining malignancy from benign breast disease with MRS alone were 83 and 87%, respectively, while a blinded MRI review reported 95 and 86%, respectively. Conclusions. Proton MR spectroscopy provides a noninvasive, biochemical measure of metabolism. The technique can be performed in less than 10 min as part of an MRI examination. MRI in combination with MRS may improve the specificity of breast MR and thereby, influence patient treatment options. This may be particularly true with less experienced breast MRI readers. In exams where MRI and MRS agree, the additional confidence measure provided by MRS may influence the course of treatment.

Introduction Upon discovery of a suspicious mass, an accurate diagnosis regarding the presence or absence malignancy is paramount. Assessment of breast tissue with imaging modalities such as mammography, and most recently magnetic resonance imaging (MRI) provide a means for lesion detection and diagnosis. Mammography provides a widely available, reliable, and costeffective screening tool. However, mammography has decreased efficacy in patients with radiographically dense breasts, patients with silicone augmentation and patients with previous surgery [1]. MRI provides

benefit not only to patients with a compromised mammography situation, but also to all patients by defining morphologic features of lesion architecture as well as dynamic contrast enhancement characteristics [2]. Both mammography and MRI provide an excellent source of diagnostic information, however, their interpretation rely on the reader’s experience [3]. Upon detecting a suspicious mass with an imaging modality, confirming the diagnosis of cancer depends upon histopathological proof. Approximately 75% of mammographically detected breast lesions in which surgical biopsy is performed ultimately prove to be benign [4]. The intricacy of tumor cell biology also

46

KM Cecil et al.

limits the standardization of tumor evaluation. Pathology sampling methods as well as descriptive errors inherent in both the radiological and histopathological techniques contribute to the failure of breast cancer identification. Additional evaluation and staging tools are necessary to improve patient management and outcome. MR spectroscopy (MRS) has been demonstrated as a useful diagnostic technique in providing metabolic information of suspicious lesions primarily in the brain. This technique has proven useful in discriminating malignancy from benign and normal tissues (for review [5]). The presence of the choline resonance from an in vivo proton MR spectroscopy assessment of breast lesions might provide a useful criterion for determining malignancy in the breast [6]. This study examined a large patient cohort with an MR imaging and spectroscopy protocol.

Materials and methods This study protocol was approved by the Institutional Review Board at the University of Pennsylvania. Informed consent was obtained from all participants in this study. Patients were referred for an MR examination with MRS upon presenting with either a palpable mass or abnormal mammographic findings. Thirtyeight patients were enrolled for this study which occurred between September 1996 and March 1998. Initially, participants for the study were selected with the criteria of a suspicious mass with a minimum size of 1 cm in diameter. In an effort to study benign processes, subsequent patients with abnormal mammographic findings without focal masses were recruited for the MRI/MRS study. Pathologic correlations were made for all patients using either excisional or needle biopsy. These studies were performed on a 1.5 T MR scanner (General Electric Signa
, Milwaukee, WI, operating with 5.5–5.7 software) using a phased array multicoil. Axial and sagittal T1 weighted spin echo (TR 500 ms/TE 16 ms), sagittal fat-saturated T2 -weighted fast spin echo (TR 5000 ms/TE 114 ms), and sagittal pre/post contrast (Gd-DTPA 0.1 mmole/kg, Magnevist, Berlex Laboratories, Wayne, NJ) enhanced fatsaturated three-dimensional (3D) fast spoiled gradient recalled echo (3D FSPGR), (TR 9.3 ms/TE 2.2 ms/flip angle 4◦ ) images using a small field of view (12– 18 cm), 512 X 256 matrix, and 2–3 mm slices were acquired from the affected breast.

Solvent suppressed and unsuppressed stimulated acquisition mode (STEAM) (TR 2000/TE 31 and/or 270 ms) localized spectra were acquired for each patient. Samples from the region of interest (voxel) were positioned by the spectroscopist (#3) in the center of a focal mass or suspicious region referred from the surgeon. The suspicious mass was determined from either a location determined from previous imaging studies, palpation with a Vitamin E capsule adhered to the breast as a marker, or by visual inspection of MR images. When multiple lesions presented, the voxel was positioned on the one noted from the surgical consultation. This was to ensure the pathology report would correlate with the findings of our MRS study. Voxel sizes are either 1 cm3 (1 cm each direction) or 3.4 cm3 (1.5 cm each direction). Select patients (17 of the 38) were sampled using both short (31 ms) and long (270 ms) echo spectroscopy. The protocol was adapted during the study from a previous report [6] to include the long echo spectroscopy and pre-contrast MRS acquisition in an effort to reduce the total examination time and avoid potential contrast agent influence of the spectral resonances. Water suppressed data were acquired for 4.5 min (128 averages) and unsuppressed data were acquired for 30 s (8 averages). The average time for combined imaging and spectroscopy was less than 55 min. Clinical diagnoses based on MRI were made using the contrast enhancement characteristics and architecture of the detected lesions. The diagnostic model used for the interpretation of the MR findings has been described previously [7]. Two radiologists provided interpretation with knowledge of the clinical histories of the patients. These reports were part of a clinical research study. A second, blinded review of the MR images, without knowledge of clinical histories, was also performed by a radiologist, different from the two clinical readers, with experience in breast MRI interpretation. The reader was asked to evaluate the images for the most suspect enhancing mass only with choices of benign, malignant or uncertain. Spectral processing for the MRS was performed by the expert spectroscopist (#3) using the SAGE/IDL software package (GE Medical Systems, Milwaukee, WI). Spectra from the four coils were combined by processing each individually and weighting each signal by its signal-to-noise ratio (SNR) using the method described by Wald et al [8]. Chemical shifts are referenced to water at 4.7 ppm. Three spectroscopists qualitatively assessed the presence of choline in a blinded review of the water

MR of breast lesions suppressed spectra. The three readers chosen for this study represent the categories described in Table 1. Each reader was given a copy of the processed (zerofilled, line broadened, Fourier transformed, chemically referenced to water for frequency placement and phased) spectra from all patients. The processed spectra used for review were those acquired at long echo time except for two patients who only had short echo data. The readers were asked to evaluate the spectra for the simple presence of choline (resonance at 3.2 ppm). The readers rated the spectra by responding yes, no or uncertain. Since the spectra were presented in a coded fashion without the MR images, the expert spectroscopist (#3) who acquired the data, had no recollection of the patient histories, voxel placement or the processed spectra. Table 1. Categories of readers for proton MR spectroscopy Reader 1

Reader 2

Reader 3

Trained spectroscopist but does not routinely read MRS data and never read breast spectra prior to this study. Trained spectroscopist who routinely reads neurospectroscopy exams, however does not routinely read breast spectra. Trained spectroscopist who routinely reads both neurological and breast spectra.

Results This study evaluated breast lesions with MR imaging and spectroscopy and correlated the diagnostic findings with pathology. Table 2 illustrates the spectroscopic acquisition parameters, the presence or absence of choline derived from a consensus of the spectroscopists, the patient demographics, the histopathology summary and the findings from the two, separate imaging reviews. An example illustrating the MRI/MRS findings of invasive ductal carcinoma upon contrast administration is shown in Figure 1. The peak at 3.2 ppm represents an elevation of composite choline compounds. However, Figure 2 shows a benign process (fibroadenoma) after administration of contrast. The corresponding proton MR spectrum indicates no evidence of choline moieties. Initial assessment of the patient MRS data sampled at short and long echo times was performed by the expert spectroscopist (#3). The review of the spectra indicated that the long echo spectroscopy provided

47

better illucidation of the choline in exams where both echotimes were sampled. The diminished signal originating from the lipid moieties due to the long echo time afforded easier detection of the choline resonance. The use of short echo spectroscopy is useful for detection of lipid involvement in many cancers. However, in the fatty tissue of the breast, lipid involvement from a malignant tumor can not be distinguished from partial voluming of the breast. Three spectroscopists evaluated processed spectra from patient studies without patient information or MRI images. The statistics for each reader and a consensus between at least two readers are summarized in Table 3. Although not considered in the statistics, we allowed the readers to categorize a spectra as uncertain. For comparison between MRI and MRS results, the consensus MRS findings indicate 19/23 confirmed cancers and 13/15 benign processes were accurately diagnosed on the basis of MRS alone (accuracy 84%, sensitivity 83% and specificity of 87%). MRS called four carcinomas benign. For the benign processes, the MRS readers reported a benign tubular adenoma and a scan of fibrocystic disease as positive for choline. MRI interpretations were read independent of the spectroscopy information content. The results of the MRI reader studies are presented in Table 4. These exams, imaged for clinical and surgical planning, were read by two radiologists whose primary research emphasis focuses on breast imaging. Thus, the reader experience level is extremely high. The MR studies were also performed after a mammographic or surgical consultation. In many exams, the reader had access to the films and reports from prior imaging studies. Screening was not the objective for these studies, but rather the analysis of architectural (morphological) features and dynamic contrast enhancement were the primary focus. With knowledge of clinical and other radiographic histories, 36 out of 38 exams were clinically interpreted in agreement with histopathological findings. The two studies with discrepancy included one invasive mammary cancer read as fibrosis (patient #11), and one benign process of fibrocystic disease interpreted as ductal carcinoma in situ (DCIS) (patient #37). For the blinded reviewer with less experience reading breast MRI, 33 exams were accurately read, three exams were misdiagnosed and two cases were uncertain. The exams misidentified by the blinded review included the two exams misread in the clinical reading along with DCIS (patient #8), a fibrocystic change (patient #31), and a hamartoma (patient #32).

48

KM Cecil et al. Table 2. Summary of MRI and MRS results #

Mass size (cm)

Voxel size (cc)

Cho

TE (ms)

MRS status acquisition

Age

Histopathology summary

Clinical MR reading

Blinded MRI reading

1 2 3

2 1.5 1.5

1 1 1

No No Yes

both both both

Post contrast Post contrast Post contrast

85 71 74

Carcinoma Carcinoma Carcinoma

malignant malignant malignant

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

3 4 2.5 2.5 2 2 2.7 × 0.2 2 4 2 1.5 2 4×4 1.5 3.5

3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4 3.4

Yes Yes Yes Yes No Yes Yes No Yes Yes Yes Yes Yes Yes Yes

both both both 270 270 both 270 270 both 270 270 270 270 270 270

Post contrast Post contrast Post contrast Post contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Prior to contrast Post contrast Prior to contrast

29 39 32 40 50 37 48 76 68 55 59 68 40 51 59

Carcinoma Carcinoma Carcinoma Carcinoma DCIS Carcinoma Carcinoma Fibrosis Carcinoma Carcinoma Carcinoma Carcinoma Carcinoma Carcinoma Carcinoma

malignant malignant malignant malignant benign malignant malignant uncertain malignant malignant malignant malignant malignant malignant malignant

19 20 21

2×3 NA 5

3.4 3.4 3.4

Yes Yes Yes

270 270 270

Prior to contrast Prior to contrast Prior to contrast

37 43 45

Carcinoma Carcinoma Carcinoma

malignant malignant malignant

22

2

3.4

Yes

270

Prior to contrast

69

Carcinoma

malignant

23 24

2 (a)

3.4 1

Yes No

270 both

Prior to contrast Post contrast

46 43

Invasive ductal CA Invasive ductal CA Invasive ductal CA & intraductal CA Invasive ductal CA Invasive ductal CA Invasive lobular CA Invasive ductal CA Intraductal CA Invasive ductal CA Invasive ductal CA Invasive mammary CA Invasive ductal CA Invasive ductal CA Invasive ductal CA Invasive ductal CA Invasive ductal CA Invasive ductal CA Invasive ductal intraductal CA Invasive ductal CA Invasive ductal CA Invasive ductal intraductal CA Invasive lobular and ductal CA Invasive ductal CA Fibrocystic changes

malignant benign

25 26 26

1 2 (a)

1 1 1

No No No

both 31 both

Post contrast Post contrast Post contrast

47 46 50

28 29 30 31

3×2 2 2 (a)

1 3.4 1 3.4

No No No No

both both both 31

Post contrast Post contrast Post contrast Post contrast

25 43 19 48

32

(a)

3.4

Yes

both

Post contrast

46

Carcinoma Fibrocystic change Fibroadenoma Fibroadenoma Regional enhancement Fibroadenoma Fibroadenoma Fibroadenoma Fibrocystic change Hamartoma

33

NA

3.4

No

both

Post contrast

73

34 35

5 2

3.4 3.4

Yes No

both 270

Prior to contrast Prior to contrast

18 47

36

1.3

3.4

No

270

Prior to contrast

38

37

1.5

3.4

No

270

Prior to contrast

62

Fibroadenoma Fibroadenoma Fibrocystic changes, hyperplasia Fibroadenoma Fibroadenoma Fibroadenoma Benign with microcalcifications Fibrocystic disease, extensive stromal Fibroadenoma Tubular adenoma Fibroadenoma, sclerosing adenosis Fibroadenoma, sclerosing Fibrocystic disease, papillomatosis

benign benign benign benign benign benign malignant uncertain

No suspicious finding Fibroadenoma Fibrocystic change Fibroadenoma

benign

benign

DCIS

malignant

benign benign

MR of breast lesions

49

Table 2. continued #

Mass size (cm)

Voxel size (cc)

Cho

TE

MRS status acquisition

Age

Histopathology summary

Clinical MR reading

Blinded MRI reading

38

(a)

3.4

No

270

Prior to contrast

65

Fibrocystic disease, apocrine metaplasia

Multiple cysts

benign

The choline presence was assigned with at least 2/3 readers indicating detection. Mass dimensions taken from clinical MRI review (a) no mass, NA not available for measurement. The echo time (TE) used for MRS acquisition was either 31 ms, 270 ms or both.

Discussion Proton MR spectroscopy provides a complementary technique to breast MRI. Currently breast MRI is not a routine screening method for breast disease, however, it is growing to serve an important role in identifying lesions and determining the extent of disease in selected patients. Although the architectural and dynamic contrast uptake criteria are the chief measures available for MRI interpretation, additional measures of metabolism are available with MR spectroscopy. As a biochemical measure of metabolism, proton MRS can detect cellular membrane turnover and proliferation by monitoring levels of a collection of chemicals with a choline base. By using single voxel proton MRS for lesion characterization, we were able to confirm a majority of the cancerous lesions found by MRI. This study replicates previous work [6] and is consistent with other preliminary studies [9]. This study of a new patient cohort differed from the previous study by evaluating MRI interpretations and changing the acquisition methodology in some patients (employed long echo and MRS acquisition before contrast injection). However, the overall MRS result was the same. The detection of elevated composite choline levels with MRS alone showed good sensitivity and specificity for determining malignancy. MRS failed to demonstrate choline elevations in four confirmed cancers. Two cancers were also missed by the blinded MRI review (patients #11 and #8). An invasive mammary cancer exam (patient #11) was technically limited by hardware failure which was demonstrated on both the MRI and MRS. Both the expert and the blinded MRI reader were unable to call this exam malignant. A DCIS study (patient #8) was acquired following an aspiration procedure. The blinded MRI reader indicated that the region of interest demonstrated recent hemorrhage and was uncertain as to diagnosis. Blood products severely interfere with the local field homogeneity, which is extremely important for successful MRS studies.

The remaining two cancers missed by MRS are discussed below while reviewing the technical limitations of MRS. In histopathologically proven benign breast disease, MRS was useful when imaging criteria indicated malignancy. In the blinded MRI review, two exams were classified as malignant. Both MRI readers indicated a false positive finding in an exam of fibrocystic disease with papillomatosis (patient #37). The second lesion was ultimately found benign but with microcalcifications (patient #31). MRS found no evidence of choline in these two patients. The presence of elevated choline in lesions that are recognized as definitely not tubular adenomas (the only known source of false MRS choline elevations in breasts) can be a powerful indicator of the malignancy of a lesion. If this well-characterized lesion is excluded by MRI, the specificity and positive predictive value for the MRS consensus of three readers improves to 93 and 95%, respectively. MRS may be able to provide an improvement in specificity which many MRI studies report deficient. The remaining false positive was a case of fibroscystic disease with extensive stromal changes (patient #32). The expert MRI reader called this as a hamartoma while the blinded MRI reader was uncertain. By incorporating MRS, which requires only 10 min, an improvement in identification of benign breast disease is demonstrated. For benign disease where MRI may be borderline, MRS may be incorporated to potentially eliminate unnecessary biopsy procedures by demonstrating the lack of choline signal. Also, where MRI interpretation alone suggests a conservative course of treatment with imaging followup recommended after a few months, MRS may provide the presence or absence of metabolic indicators for a diagnosis. Our study demonstrated that the MRI readers had outstanding accuracy, sensitivity and specificity in determining malignancy. MRS is at the lower limit of its

50 KM Cecil et al. Figure 1. Provides an example of a case with invasive ductal carcinoma (patient #4). The figure shows the image taken after the administration contrast agent along with the MR spectrum. Note the appearance of a resonance at 3.2 ppm corresponding to elevations in composite choline. Lipid signals are also present in spectra between 0.8 ppm and 1.5 ppm.

MR of breast lesions utility and may only be of confirmatory value to such readers. For the original MRI interpretation model used in the MRI reviews, a specificity of 69% was reported [7]. Other reviews of breast MRI show good to excellent sensitivity with variable specificity depending upon the diagnostic criteria (e.g. architecture, enhancement rates and curve types) [10–12]. The influence of prior mammographic examinations as well as clinical histories is partially accounted for in this study. Prior reports of the breast interpretation model used in this study indicated that knowledge of clinical histories did not statistically change the reader performance statistics [2, 7]. However in this small study, the specificity and positive predictive value are weakened without the knowledge of clinical histories. The influence of the exams called uncertain is also demonstrated in the statistics (Table 4). The blinded reviewer called uncertain both a benign and a malignant process. If forced into a diagnosis, the reviewer would have called the two processes benign. The resulting statistics are shown as the (b) condition of Table 4 with a loss of sensitivity and negative predictive value. Although there is some difference from the naïve to the experienced spectroscopist, the sensitivity and the specificity are fairly high for the spectroscopy readers (kappa statistic between expert and novice 0.787). Readers 2 and 3 indicate that an experience threshold has been surpassed. Upon crossing the experience threshold, there was no difference between a reader who occasionally reads breast spectra and one who reads spectra weekly (kappa statistic 1.00). These readings are less rigorous than a review of a MRI exam. Single voxel spectroscopy can not aid in the detection of suspicious lesions; however, it can play a role in the characterization of lesions for a typical radiologist by improving specificity. Using a linear combination of the MRI and MRS findings, specificity can be improved with a sacrifice in sensitivity. Unfortunately, there may exist selection bias in our population since the patients originated primarily from a surgical service. Another source of selection bias is the presence of enhancing lesions in a majority of the patients. The incidence of malignancies in this population (23/38) is much greater than what is reported in MRI studies as well as in the general population. In some patients, multiple lesions were found, but MRS was typically performed on a single lesion. Single voxel MRS requires easily detected lesions, a marked location of palpable mass or the coordinates from a previ-

51

ous mammographic or MRI exam. In general, this requirement remains a significant challenge, which may preclude other sites from reproducing our results. Potential errors may also arise by the changing inclusion criteria and sampling techniques (Table 2). Although the inclusion criteria expanded to include patients without focal masses (5/38), no false negative MRS examination was reported among this subset. A false positive MRS result was found in Patient #32 described previously. Thus, we believe there was limited error because of this change. As noted previously, two MRS acquisitions were performed in 17/38 patients using short and long echo MRS techniques. The MRS reader review of patient spectra included the long echo acquistion only, except for the two patients did not have successful acquisition of long echo MRS. The expert and trained spectroscopist MRS readings in these two cases (Table 2) were correct upon comparison with pathology. The novice MRS reader called one spectra uncertain. An equivalent number of errors, either false positive or false negative, were found in the MRS readings when performed prior to or post contrast acquisition in both the malignant and benign pathologies (Table 2). The influence of this change is potentially the greatest by introducing sampling errors in the acquisition phase of the study. In a previous report, [6] spectroscopy was acquired after contrast injection for more accurate voxel placement. In a clinical setting, the region of interest may not be known in advance. Again, our patients had detailed clinical histories with the region of interest well documented in the surgical consultation. Since our goal was to match the MRS acquisition with the biopsy site, the expert spectroscopist had to feel confident enough to select this region from the non-contrast images. On the other hand, alterations of the metabolite signals have been suggested in postcontrast acquired MRS, specifically, that the loss of choline signal upon MRS sampling post contrast agent injection [13]. The broadening of the choline metabolite peak in spectra acquired post contrast acquisition may make discernment of the choline more difficult. For this study, we began to sample for MRS prior to contrast administration (Table 1) until our expertise level was shown reliable for identifying the questioned lesion. It evolved such that MRS sampling after contrast administration was only performed if the questioned lesion was not recognized with non-contrast MRI. It was also a technical concern that prompted this protocol switch. Reconstruction of 3D FSPGR images

52

KM Cecil et al.

Figure 2.

MR of breast lesions Table 3. Statistics for the blinded MRS reader review MRS reader

A

S

P

PPV

NPV

U-CA

U-B

1 2 3 Consensus

76 84 86 84

82 83 86 83

67 86 87 87

82 90 90 90

67 75 81 76

1 0 1 0

3 1 0 0

The table illustrates the statistical differences among MRS readers. A = accuracy, S = sensitivity, P = specificity, PPV = positive predictive value, NPV = negative predictive value. The cases regarded as uncertain are excluded from the statistics, but are shown in the table as U-CA (pathology proven cancer) and U-B (pathology proven benign). MR Spectra were read without MR images or knowledge of clinical histories. Table 4. Results of MRI reader study MRI reader

A

S

P

PPV

NPV

U-CA

U-B

Clinical Blinded (a) Blinded (b) Blinded (c)

95 92 89 89

96 95 91 96

93 86 87 80

96 91 91 88

93 92 87 92

1 0 0

1 0 0

Statistics for MRI reader studies. (a) Uncertain findings are excluded from statistics. (b) Uncertains are considered benign. (c) Uncertains are considered malignant.

upon contrast administration precluded the acquisition of MRS because use of the image processing capabilities of the MR scanner. The patient would wait several minutes before MRS acquisition and would become more restless. This could potentially introduce patient motion resulting in the voxel placement as a source of error for the cases with false negatives. Limited signal to noise of the spectra in voxels with only 1cc volume may also contribute to the false negatives. Partial voluming with the fatty breast tissue could potentially mask the choline signal in such a small sample. In our malignant population, the false negatives for patients #1 and 2 could have ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 2. Provides an example of a case with a fibroadenoma (patient #30). The figure shows the fibroadenoma (adjacent to the skin below the nipple) taken after the administration contrast agent along with the MR spectrum. The MR spectrum is zoomed into to demonstrate the absence of the choline signal. Lipid signals are present in spectra between 2.0 and 2.5 ppm. Additional lipid peaks are excluded due to their overwhelming signal intensity.

53

some measure of error attributed to the post acquisition status. If the these two cases are excluded, our MRS findings would approach the blinded MRI findings in histopathologically proven cancers described earlier. Thus, if this work was replicated in another patient population, the sensitivity statistics could potentially be equivalent to the MRI statistics. At present, breast MRS remains more technically demanding than MRI. However, the technical demands of proton MRS are not prohibitive. With the advent of automatic brain proton spectroscopy software packages, spectroscopy sequences are added on to routine neurological MRI examinations in the United States. Automation for breast MRS could be achieved if warranted. Proton MR spectroscopy of the breast provides a metabolic index of cellular proliferation useful in aiding the specificity of MRI. MRS findings may aid in eliminating biopsies or surgical procedures in cases of benign breast disease. Patients with MRI and MRS evidence of cancer may receive more rapid and aggressive treatments without additional biopsy confirmation and potential seeding. MRS findings may change the course of treatment in both patient groups. Nationally, as more MR scanners are equipped with spectroscopy capabilities, there will be more sites without dedicated spectroscopists and physicists. As these sites incorporate spectroscopy into clinical MR examinations, further training of radiologists to read breast MRS could potentially increase the specificity and diagnostic accuracy of breast disease. Further development of multivoxel spectroscopic methods, referred to as chemical shift imaging (CSI), may improve the sensitivity of choline detection. However, until such a development, single voxel proton MR spectroscopy can complement lesion characterization along with structural and dynamic MRI. These results suggest that a multicenter trial involving single voxel proton MRS of the breast are appropriate. Acknowledgements This work was presented in part as a poster presentation at the Sixth Scientific Meeting and Exhibition of the International Society for Magnetic Resonance in Medicine, Sydney, Australia, April, 1998 and at the 85th Scientific Assembly and Annual Meeting of the Radiological Society of North America held December 1999 in Chicago, Illinois. We thank Norman Butler, Tanya Kurtz, Doris CainEdwards, M. Robert Willcott and Jean McDermott for

54

KM Cecil et al.

their assistance. This work was supported by grants RO1-CA58358 (NIH), RR 02035 (NIH), DAMD1793-C-3086 (U.S. ARMY), and T32-HL-07614-06A1 (NIH).

8.

9.

References 1. Friedrich M: MRI of the breast: state of the art. Eur Radiol 8(5): 707–725, 1998 2. Nunes LW, Schnall MD, Siegelman ES, Langlotz CP, Orel SG, Sullivan D, Muenz LA, Reynolds CA, Torosian MH: Diagnostic performance characteristics of architectural features revealed by high spatial-resolution MR imaging of the breast. AJR Am J Roentgenol 169(2): 409–415, 1997 3. Schmidt F, Hartwagner KA, Spork EB, Groell R: Medical audit after 26,711 breast imaging studies: improved rate of detection of small breast carcinomas. Cancer 83(12): 2516– 2520, 1998 4. Kopans DB: Preoperative imaging-guided needle placement and localization of clinically occult lesions. In: Kopans DB (ed) Breast Imaging. Philadelphia, Lippincott, 1989, pp 320– 341 5. Negendank W: Studies of human tumors by MRS: a review. NMR Biomed 5(5): 303–324, 1992 6. Roebuck JR, Cecil KM, Schnall MD, Lenkinski RE: Human breast lesions: characterization with proton MR spectroscopy. Radiology 209(1): 269–275, 1998 7. Nunes LW, Schnall MD, Orel SG, Hochman MG, Langlotz CP, Reynolds CA, Torosian MH: Breast MR imaging: inter-

10.

11.

12.

13.

pretation model [see comments]. Radiology 202(3): 833–841, 1997 Wald LL, Moyher SE, Day MR, Nelson SJ, Vigneron DB: Proton spectroscopic imaging of the human brain using phased array detectors. Magn Reson Med 34(3): 440–445, 1995 Gribbestad IS, Singstad TE, Nilsen G, Fjosne HE, Engan T, Haugen OA, Rinck PA: In vivo 1H MRS of normal breast and breast tumors using a dedicated double breast coil. J Magn Reson Imaging 8(6): 1191–1197, 1998 Stomper PC, Herman S, Klippenstein DL, Winston JS, Edge SB, Arredondo MA, Mazurchuk RV, Blumenson LE: Suspect breast lesions: findings at dynamic gadolinium-enhanced MR imaging correlated with mammographic and pathologic features. Radiology 197(2): 387–395, 1995 Heiberg EV, Perman WH, Herrmann VM, Janney CG: Dynamic sequential 3D gadolinium-enhanced MRI of the whole breast. Magn Reson Imaging 14(4): 337–348, 1996 Helbich TH, Becherer A, Trattnig S, Leitha T, Kelkar P, Seifert M, Gnant M, Staudenherz A, Rudas M, Wolf G, Mostbeck GH: Differentiation of benign and malignant breast lesions: MR imaging versus Tc-99m sestamibi scintimammography. Radiology 202(2): 421–429, 1997 Lenkinski RE, Cecil KM, Desiderio L, Hackney DB: Gadolinium does not change the Diagnosis of Brain Tumors using Proton MRS. Radiology 209P: 237, 1998

Address for offprints and correspondence: Kim M. Cecil, Children’s Hospital Medical Center, Imaging Research Center, Department of Radiology, 3333 Burnet Avenue, Cincinnati, OH 45229-3039; Tel.: 513-636-8559; Fax 513-636-3754; E-mail: [email protected]