AJR:181, November 2003

1211

Comparison of High-Field-Strength Versus Low-Field-Strength MRI of the Shoulder

OBJECTIVE.

Previous studies have reported similar results of shoulder MRI versus ar-throscopy for high-field-strength (1.5-T) and low-field-strength (0.2-T) units. We report our ex-perience with the accuracy of high- versus low-field-strength units versus arthroscopy fordetection of supraspinatus tendon tears and labral tears in the same patients.

SUBJECTS AND METHODS.

Three musculoskeletal radiologists prospectively inter-preted shoulder MRIs from 40 patients who had a complete shoulder MRI examination on a0.2-T system and limited imaging on a 1.5-T unit. Proton-density axial and fat-saturated T2-weighted coronal and sagittal sequences were performed. Each radiologist interpreted theopen unit images first and the high-field-strength images second. Results from 28 patientswho also underwent arthroscopy were also compared with the MRI interpretations. All scanswere then retrospectively reinterpreted by consensus of the three reviewers, who were un-aware of the patient’s name, results of previous MRI, or arthroscopy report (if any).

RESULTS.

High-field-strength images altered reviewers’ interpretations of low-field-strengthscans for nine of 40 patients. In four patients, full-thickness supraspinatus tendon tears could bediagnosed definitively on the high-field-strength unit but not on the open unit. Three labral tearsand two superior labral anteroposterior lesions could be depicted definitively on the high-field-strength unit but not on the open unit. All tears were confirmed at arthroscopy.

CONCLUSION.

High-field-strength MRI units provide better spatial and contrast reso-lution and allow more accurate interpretations than low-field-strength units; these findingsmay affect clinical treatment.

everal previous studies have com-pared MRI of the shoulder fromhigh-field-strength (1.5 T) versus

low-field-strength (0.2–0.3 T) imaging units.Most of those previous studies indicated nosignificant clinical differences in interpreta-tions of low- versus high-field-strength units.Such studies were limited in that the same pa-tients were not imaged on both high- and low-field-strength units. Also, most of these studieswere retrospective in nature [1–4].

Open low-field-strength MRI units are pop-ular imaging tools because claustrophobic pa-tients tolerate them better than the moreconfining high-field-strength units. Someobese patients fit only into open units. Mostlow-field-strength units also cost less to pur-chase and maintain than high-field-strengthunits. However, often the low-field-strengthunits do not have fat suppression, and imagingtakes considerably longer. The additional im-

aging time increases the risk of patient motionand resultant motion artifacts.

The strength of the magnetic field is mea-sured in gauss or tesla units (10,000 G = 1 T).In general, higher field strength improves thesignal-to-noise ratio (SNR). SNR, contrast,and resolution increase almost linearly withfield strength, at least up to 1.5 T. The in-creases allow faster scanning and lower the in-cidence of motion artifacts [5, 6].

Image noise can affect visualization of low-contrast lesions. The ratio of signal difference(or contrast) to noise indicates how well a lesioncan be detected. Faster scanning times and thin-ner slices tend to reduce noise.

Image noise ad-versely affects the reviewer’s ability to detectlow-contrast lesions. Noise is reduced relative tosignal by increasing voxel volume. Voxel vol-ume can be increased by increasing the field ofview or increasing the slice thickness.

However,higher voxel volume may cause volume-averag-

Thomas Magee

1

Marc ShapiroDavid Williams

Received January 17, 2003; accepted after revision May 14, 2003.

1

All authors: Department of Radiology, Neuroimaging Institute, 27 E Hibiscus Blvd., Melbourne, FL 32901. Address correspondence to T. Magee (tmageerad@cfl.rr.com).

AJR

2003;181:1211–1215

0361–803X/03/1815–1211

© American Roentgen Ray Society

S

1212

AJR:181, November 2003

Magee et al.

ing signal abnormalities that make lesions moredifficult to detect [5, 6].

We undertook this study to determinewhether imaging the same patients on bothhigh- (1.5-T) and low-field-strength (0.2-T)units would affect MRI interpretations, and ifso, whether these differences in interpretationwould affect clinical treatment.

Subjects and Methods

One hundred fifty-three consecutive patients un-dergoing open MRI were asked if they would agreeto additional limited imaging on an adjacent 1.5-TMRI unit. Those who agreed to the additional proce-dure were scanned on the same day as their low-field-strength scan. Most of the 113 patients who didnot wish to have additional imaging on the high-field-strength unit cited time constraints as the reasonfor declining additional imaging (e.g., the need to re-turn to work, perform errands). Thirty-two of the pa-tients who refused additional imaging citedclaustrophobia as the reason that they did not wishadditional imaging. Patients ranged in age from 21 to74 years (mean, 47 years).

All patients had a complete shoulder MRI exami-nation on an open 0.2-T system (General ElectricMedical Systems, Milwaukee, WI). MRI sequenceson the open system were as follows: coronal T1-weighted images (TR/TE, 685/12; imaging time, 6min 43 sec), T2-weighted images (3,350/94.5; imag-ing time, 7 min 21 sec), and STIR images (2,500/22;imaging time, 6 min 46 sec; inversion time, 75 msec)with a 20-cm field of view, 4-mm slice thickness (forT1- and T2-weighted images), 5-mm slice thickness(for STIR images), and a net excitation of 4 for T1-and T2-weighted images and of 5 for STIR images.

Sagittal STIR images (2,500/22; imaging time, 6 min46 sec; inversion time, 75 msec) with a 5-mm slicethickness, net excitation of 5, and a 20-cm field ofview were obtained. Axial gradient-echo images(685/20; imaging time, 7 min 21 sec) with a 20-cmfield of view, a net excitation of 4, slice thickness of 4mm, and a flip angle of 20° were obtained. Some pa-tients had axial T1-weighted images (685/12; imagingtime, 6 min 43 sec) and T2-weighted images (3,350/94.5; imaging time, 7 min 41 sec) with a net excita-tion of 4 and an 18-cm field of view. Some patientswere not imaged in all sequences.

Forty patients agreed to undergo the additionalMRI on the adjacent high-field-strength (1.5-T) scan-ner (General Electric Medical Systems). Coronal andsagittal fast spin-echo T2-weighted images (2,500/68;imaging time, 2 min 27 sec for each sequence) with a14-cm field of view, 4-mm slice thickness with 1-mminterslice gap, and a net excitation of 3 were obtained.Patients also underwent axial fast spin-echo proton-density images (2,375/34; imaging time, 2 min 36sec) with a 14-cm field of view, 4-mm slice thicknesswith a 1-mm interslice gap, and a net excitation of 2.A phased array shoulder coil was used on all patients.Three experienced musculoskeletal radiologists pro-spectively reviewed the shoulder scans from the 40patients who agreed to undergo both series of MRIs.Each radiologist interpreted the low-field-strength im-ages first and the high-field-strength images secondand determined whether any differences betweenlow-field-strength and high-field-strength scans al-tered their interpretation. Twenty-eight patients alsoagreed to arthroscopy, and the results were comparedwith the prospective interpretations of the high- andlow-field-strength units. All arthroscopies were per-formed less than 2 months after the MRI.

On prospective review, MRIs from both units wereanalyzed by one radiologist for the presence or ab-

sence of full-thickness supraspinatus tendon tears, su-perior labral anteroposterior (SLAP) tears, andanterior or posterior labral tears. Full-thickness su-praspinatus tendon tears were defined as areas of fluidsignal traversing the entire thickness of the su-praspinatus tendon on T2-weighted or STIR images,or as supraspinatus tendon retraction. SLAP tearswere defined as areas of high signal on T2-weightedor STIR images in the superior labrum or contour ir-regularities of the superior labrum. Anterior or poste-rior labral tears were defined as areas of abnormalincreased signal on T2 or STIR images or contour ir-regularities of the anterior or posterior labrum.

All 40 MRIs were retrospectively reinterpreted byconsensus of the three radiologists, who were un-aware of patient’s name, results of previous MRIs, orarthroscopy report (if any). The reinterpretations weredone at one sitting, with high- and low-field-strengthimages from the same patient presented out of orderin a random fashion so as not to bias the reviewers.

Results

Forty patients were scanned on both high-(1.5-T) and low-field-strength (0.2-T) units. For31 patients, prospective reviewer interpretationswere the same for both units. Arthroscopic re-sults were available for 19 patients. Sixteen ofthe 19 showed full-thickness supraspinatus tearson both low- and high-field-strength prospectiveMRI interpretations. Two patients had anteriorlabral tears and one had a SLAP lesion on bothhigh- and low-field-strength prospective MRI in-terpretations. All MRI reviews were confirmedarthroscopically. Partial-thickness tears and ten-dinosis were not included in the results becausethese findings did not affect surgical treatment.

A B C

Fig. 1.—47-year-old man with shoulder pain.A, T1-weighted axial 0.2-T image (TR/TE, 685/12) shows no abnormality in posterior labrum (arrow).B, STIR axial 0.2-T image (2,500/22; inversion time, 75 msec) shows no definite abnormality in posterior labrum. Subtle area of increased signal (arrow) may be visible inposterior labrum. C, T2-weighted axial 1.5-T image (2,500/68) shows abnormal signal and irregularity in posterior labrum confirmed at arthroscopy to be posterior labral tear (arrow).

High- Versus Low-Field-Strength MRI of the Shoulder

AJR:181, November 2003

1213

Retrospective blinded consensus interpretationsby the three radiologists did not differ from pro-spective MRI reviews.

High-field-strength images altered prospec-tive interpretations for nine patients. Two ante-rior labral tears, a posterior labral tear, and twoSLAP lesions could be definitively seen on thehigh-field-strength unit but not on the low-field-strength unit (Figs. 1–3).

The anterior la-bral tears were in the anteroinferior portion ofthe labrum and both were detached. The poste-rior labral tear was not detached. These labraltears were confirmed at arthroscopy.

In four patients, it was difficult to determineon the open MRI whether a partial- or full-thickness supraspinatus tendon tear was present.The high-field-strength MRI unit allowed defin-itive interpretation of a full-thickness tear and amore accurate delineation of the degree of ten-

don retraction. Three of these full-thickness su-praspinatus tendon tears were small tearswithout any appreciable degree of tendon re-traction (Fig. 4). One of the four full-thicknesssupraspinatus tendon tears was retracted by 3cm (Fig. 5). Fibrosis in the area of the tear pre-vented it from showing high signal and made itdifficult to see the free edge of the retracted su-praspinatus tendon on the images obtained onthe low-field-strength unit. These full-thicknesstears were all confirmed at arthroscopy.

Discussion

The high-field-strength MRI altered re-viewer interpretation in nine of 40 cases. Allthe high-field-strength interpretations in thosecases were confirmed at arthroscopy. Our find-ings differ from those of most previous studies

comparing high- and low-field-strength MRIinterpretations. Shellock et al. [1] comparedinterpretations in 47 patients who underwentMRI in a 0.2-T extremity MRI system withfindings at arthroscopy. They found that thesensitivity, specificity, and predictive valuesfor MRI of rotator cuff and labral tears com-pared favorably with those values published inthe literature for mid- and high-field-strengthMRI systems. Their study differs from ours inthat they did not scan the same patients onboth high- and low-field-strength units andcompare the results. They compared the resultsof one set of patients on a low-field-strengthunit with results from other sets of patient re-sults previously published in the literature.

Tung et al. [2] compared high-field-strengthversus low-field-strength MRI for the diagno-sis of SLAP lesions. They found high-field-

A B C

Fig. 2.—56-year-old man with shoulder pain.A, T2-weighted coronal 0.2-T image (TR/TE, 3,350/94.5) shows no abnormality in superior labrum (long arrow) and suspected partial-thickness supraspinatus tear (shortarrow).B, STIR coronal 0.2-T image (2,500/22; inversion time, 75 msec) shows no abnormality in superior labrum (long arrow) and depicts partial-thickness supraspinatus tendontear (short arrow) in articular surface.C, T2-weighted coronal 1.5-T image (2,500/68) shows superior labral anteroposterior tear (long arrow) and partial-thickness supraspinatus tendon tear in articular surface(short arrow), that were confirmed at arthroscopy.

A B

Fig. 3.—38-year-old woman with shoulder pain.A, T2-weighted coronal 0.2-T image (TR/TE, 3,350/94.5)shows suspected abnormality in superior labrum (ar-row). STIR coronal imaging was attempted, but patientcould not remain motionless long enough.B, T2-weighted coronal 1.5-T image (2,500/68) showsdefinite superior labral anteroposterior tear (arrow),that was confirmed at arthroscopy.

1214

AJR:181, November 2003

Magee et al.

strength imaging to be better than low-field-strength imaging for the diagnosis of SLAP le-sions. The study by Tung et al. compared resultsfrom one set of patients scanned on a low-field-strength unit with those from another set of pa-tients scanned on a high-field-strength unit.

Other studies have compared low- and high-field-strength units with two sets of patients.These other studies found no significant differ-ences in MRI interpretation between the twosets [3, 4].

Our study differs from previous studies inthat it is prospective. Also, all patients in thestudy underwent MRI on both high- and low-

field-strength units. The low-field-strengthinterpretations were provided first and thehigh-field-strength interpretations second.The radiologists then determined whether theirinterpretations had changed.

A difference in interpretation does not neces-sarily affect clinical treatment. In the nine in-stances of differences in reviewer interpretationbetween the high- and low-field-strength units,the orthopedic surgeon involved indicated thatthe high-field-strength interpretations did alterclinical treatment because surgery was per-formed to address the findings shown on thehigh-field-strength unit. All nine of these

high-field-strength interpretations were con-firmed arthroscopically.

High-field-strength units provide more sig-nal to noise than low-field-strength units. Also,scans are performed much more quickly.Therefore, patient motion is less on the high-field-strength units than on low-field-strengthunits [5, 6].

Image quality is based on contrast resolu-tion and noise. Noise is caused by randommotion of electrons produced primarily bythe coil and the patient’s body. Increasing thetime duration for signal acquisition reducesimage noise. All things being equal, the

A B C

Fig. 4.—53-year-old man with shoulder pain.A and B, STIR coronal (A) and sagittal (B) 0.2-T images (TR/TE, 2,500/22; inversion time, 75 msec) show abnormal signal (arrow) in distal supraspinatus tendon but no full-thickness tear. C, T2-weighted coronal 1.5-T image (2,500/68) shows small full-thickness supraspinatus tendon tear (arrow), that was confirmed at arthroscopy.

A B C

Fig. 5.—68-year-old man with shoulder pain.A, T2-weighted coronal 0.2-T image (TR/TE, 3,350/94.5) shows attenuation of supraspinatus tendon (arrow) but no definite full-thickness tear.B and C, T2-weighted coronal (B) and sagittal (C) images show full-thickness supraspinatus tendon tears, one with retraction (arrow, B) and the other (arrow, C) confirmedat arthroscopy.

High- Versus Low-Field-Strength MRI of the Shoulder

AJR:181, November 2003

1215

strength of signal from a voxel is propor-tional to the voxel volume. The SNR is pro-portional to the voxel volume times thesquare root of the signal acquisition time.Any technique that increases signal acquisi-tion time will increase the SNR. Any tech-nique that decreases voxel volume in anattempt to improve spatial resolution will de-crease the SNR. The spatial resolution re-flects the sharpness of the image [5, 6].

High-field-strength units inherently pro-vide a better SNR than low-field-strengthunits. They allow faster imaging and thinnerslice thicknesses. The decrease in slicethickness (voxel volume) allows better spa-tial resolution [5, 6].

Thinner slice thickness (4 vs 5 mm on thelow-field-strength unit) and a smaller fieldof view (14 cm on high-field-strength unit vs20 cm on the low-field-strength unit) wereused for the high-field-strength unit images.This most likely resulted in better visualiza-

tion of subtle signal abnormalities, allowingbetter resolution of small full-thickness tears,supraspinatus tears, small labral tears, andsubtle SLAP lesions. These small tears didnot resolve as well on the low-field-strengthunits because the larger slice thicknessneeded to maintain adequate SNR requiredmore volume averaging.

A potential limitation of this study is that allmusculoskeletal radiologists were from thesame institution and performed consensus in-terpretation on retrospective review. These in-dividuals may interpret MRIs more uniformlythan individuals from different institutions.

In conclusion, the superior spatial andcontrast resolution of the high-field-strengthunits may result in more accurate interpreta-tion of full-thickness supraspinatus tendontears and labral tears in some patients thanwould be possible with low-field-strengthunits. The changes in reviewer interpretationmay have an effect on clinical treatment.

References

1. Shellock FG, Bert JM, Fritts HM, et al. Evalua-tion of the rotator cuff and glenoid labrum using a0.2 tesla extremity magnetic resonance (MR) sys-tem: MR results compared to surgical findings.

JMagn Reson Imaging

2001;14:763–7702. Tung GA, Entzian D, Green A, Brody JM. High-

field and low-field MR imaging of superior gle-noid labral tears and associated tendon injuries.

AJR

2000;174:1107–11143. Allmann KH, Walter O, Laubenberger J, et al.

Magnetic resonance diagnosis of the anterior la-brum and capsule: effect of field strength on effi-cacy.

Invest Radiol

1998;33:415–4204. Shih TT, Chen WG, Su CT, et al. MR patterns of

rotator cuff and labral lesions: comparison be-tween low-field and high-field images.

J FormosMed Assoc

1993;92:146–1515. Joseph PM. Principles of image formation. In:

Atlas SW, ed.

Magnetic resonance imaging of thebrain and spine

, 2nd ed. Philadelphia: Lippincott-Raven, 1996:46–63

6. Hendrick RE. Image contrast and noise. In: StarkDD, Bradley WG, eds.

Magnetic resonance imag-ing

, 3rd ed. St. Louis: Mosby, 1999:43–67