FULL PAPER

Comparison of Two Ultrashort Echo Time Sequences forthe Quantification of T1 Within Phantom and HumanAchilles Tendon at 3 T

Peter Wright,1* Vladimir Jellus,2 Dennis McGonagle,3 Matthew Robson,4

John Ridgeway,1 and Richard Hodgson3

Ultrashort echo time (UTE) techniques enable direct imaging ofmusculoskeletal tissues with short T2 allowing measurementof T1 relaxation times. This article presents comparison of opti-mized 3D variable flip angle UTE (VFA-UTE) and 2D saturationrecovery UTE (SR-UTE) sequences to quantify T1 in agar phan-toms and human Achilles tendon. Achilles tendon T1 values forasymptomatic volunteers were compared to Achilles tendon T1

values calculated from patients with clinical diagnoses ofspondyloarthritis (SpA) and Achilles tendinopathy using anoptimized VFA-UTE sequence. T1 values from phantom data forVFA- and SR-UTE compare well against calculated T1 valuesfrom an assumed gold standard inversion recovery spin echosequence. Mean T1 values in asymptomatic Achilles tendonwere found to be 725 6 42 ms and 698 6 54 ms for SR- andVFA-UTE, respectively. The patient group mean T1 value forAchilles tendon was found to be 957 6 173 ms (P < 0.05) usingan optimized VFA-UTE sequence with pulse repetition time of 6ms and flip angles 4, 19, and 24�, taking a total 9 min acquisitiontime. The VFA-UTE technique appears clinically feasible forquantifying T1 in Achilles tendon. T1 measurements offer poten-tial for detecting changes in Achilles tendon due to SpA withoutneed for intravenous contrast agents. Magn Reson Med000:000–000, 2012.VC 2012 Wiley Periodicals, Inc.

Key words: T1 quantification; UTE-MRI; achilles tendon;musculoskeletal

The early phases of rheumatic diseases may preferentiallyafflict different joint components including the synovium,bone, cartilage, ligament, tendons, and entheses. Thecapabilities of magnetic resonance imaging are limited ina number of musculoskeletal tissues such as ligaments,tendon, and bone, which contain a majority of specieswith short T2 time constants, so that little or no signal isobtained from these tissues when using conventionalimaging techniques. Tendons, especially the componentsnear the enthesis, are commonly involved in diseases

such as spondyloarthritis (SpA) and osteoarthritis (1) butare difficult to assess due to their low signal.

The recent development of ultrashort echo time (UTE)imaging techniques (2,3) with effective echo times (TEs)up to 20 times shorter than those currently available onconventional clinical scanners allow signal to be obtainedfrom these short T2 tissues. As well as allowing direct vis-ualization of tendons, UTE sequences also provide thepotential to make quantitative measurements of parame-ters including the longitudinal recovery time constant, T1.T1 values of the Achilles tendon have previously beenreported in cadaveric specimens at 3 T (4,5) using UTEtechniques and at 1.5 T in volunteers (6) and patientsdiagnosed with chronic mechanical or degenerative Achil-les tendinosis using magic angle imaging techniques (7).

The aims of this study were (i) to compare two sequen-ces, saturation recovery UTE (SR-UTE) and variableflip angle UTE (VFA-UTE), against an assumed gold stand-ard inversion recovery spin echo (IR-SE) sequence formeasuring T1 in phantoms, (ii) to compare T1 values calcu-lated from the SR- and VFA-UTE sequences of the Achillestendon of healthy volunteers, and (iii) to compare Achillestendon T1 values from SpA patients with those fromhealthy volunteers using an optimized VFA-UTE sequence.

MATERIALS AND METHODS

Simulations and Optimization

The spoiled gradient echo VFA-UTE sequence was simu-lated using Eq. 1, where the signal, S, for a particularflip angle, b, can be expressed as:

S ðbÞ ¼ S0sinðbÞð1� expð�TR=T1ÞÞð1� cosðbÞexpð�TR=T1ÞÞ ½1�

where S0 is the equilibrium signal and TR is the repeti-tion time. Monte-Carlo simulations were carried out forthe VFA-UTE sequence to optimize the choice of flipangles to reduce total scan time required while maintain-ing reliable T1 values assessed by determining how farthe T1 value deviated from the model T1 value and itsassociated error in multiple model fitting simulations.Combinations of two and three flip angles from 1 to 28�

and repetition times of 6 and 8 ms were simulated.Data were simulated with model parameters of T1 ¼

600 ms, based on the Achilles tendon T1 values previouslyreported at 3 T in cadavers (4,5) and S0 ¼ 1. Randomgaussian noise was added at 0.5% S0, and the resulting

1LMBRU, Leeds Teaching Hospitals NHS Trust, Leeds, West Yorkshire,United Kingdom.2Healthcare Sector, Siemens AG, Erlangen, Germany.3Leeds Musculoskeletal Biomedical Research Unit (LMBRU), University ofLeeds, Leeds, West Yorkshire, United Kingdom.4University of Oxford Centre for Clinical Magnetic Resonance Research(OCMR), John Radcliffe Hospital, Oxford, United Kingdom.

Grant sponsors: Arthritis Research (AR) UK, National Institute for HealthResearch (NIHR).

*Correspondence to: Dr. Peter Wright, M.Sci., M.Sc., Ph.D., LMBRU,Chapel Allerton Hospital, Harehills Lane, Chapel Allerton, Leeds, WestYorkshire LS7 4SA, United Kingdom. E-mail: pjw@medphysics.leeds.ac.uk

Received 1 September 2011; revised 3 November 2011; accepted 30November 2011.

DOI 10.1002/mrm.24130Published online in Wiley Online Library (wileyonlinelibrary.com).

Magnetic Resonance in Medicine 000:000–000 (2012)

VC 2012 Wiley Periodicals, Inc. 1

simulated data fitted using an in-house program writtenin Matlab. Random noise data were generated using afunction incorporating the Ziggurat method for normaldistribution (8). The process was repeated 1000 times toproduce 1000 fitted data sets for each combination of flipangles and the mean T1 and standard deviation calculatedfor each combination. The optimized flip angle combina-tion for the VFA-UTE sequence was chosen as having thesmallest standard deviation from the fitted T1.

Phantom Data Acquisition

All MRI scanning was carried out on a Siemens 3 T Veriosystem using a 4-cm loop receive coil. Measurements weremade with IR-SE, VFA-UTE, and SR-UTE sequences.

The parameters for the IR-SE sequence were TE¼ 7.9ms,TR ¼ 10 s, acquisition matrix size ¼ 512 � 256, voxelsize ¼ 0.352 � 0.352 � 3 mm3, and bandwidth ¼305 Hz/pixel. Data were acquired using 10 differentinversion times (TIs) of 50, 100, 200, 400, 600, 800,1000, 1200, 1500, and 2000 ms chosen in random order.Only one slice was acquired to avoid errors from slicecross-talk resulting from imperfect slice profiles. Thisresulted in a total scan duration per TI of 42min 50 s.

The VFA-UTE was a 3D sequence with acquisitionparameters TE ¼ 0.07 ms, acquisition matrix size ¼ 256 �256 � 256, radial steps ¼ 30,000, voxel size ¼ 0.625 �0.625 � 0.625 mm3, bandwidth ¼ 630 Hz/pixel, 3D radialsampling, and an excitation radiofrequency pulse durationof 60 ms. Data were acquired using 26 flip angles between 1and 26� in steps of 1�, where higher flip angles were notpossible due to the short time available for the excitationpulse to occur within the sequence. A TR of 6 ms was usedfor phantom data acquisition, and TRs of 6 and 8 ms wereused for the patient and volunteer groups, resulting in atotal scan time per flip angle of 3 and 4 min respectively.

Data were acquired using seven saturation times (TS) of50, 100, 200, 400, 800, 1200, and 1600 ms for the singleslice SR-UTE sequence where the saturation radiofre-quency pulse had a duration of 1 ms. Parameters were TE¼ 0.07 ms, TR ¼ 2 s þ TS, excitation flip angle ¼ 90�,acquisition matrix size ¼ 128 � 128, voxel size ¼ 0.82 �0.82 � 3 mm3, and bandwidth ¼ 340 Hz/pixel, whichresulted in a total scan time per TS from 8 min 58 s to15 min 22 s, increasing in time with longer saturation times.

Four selected gadolinium doped agar tubes from a Euro-spin TO5 phantom (9) with a range of T1 values between370 and 960 ms (chosen to match the range observedwithin the musculoskeletal system) were imaged usingthe IR-SE, VFA-UTE, and SR-UTE sequences.

In Vivo Studies

Eight healthy asymptomatic volunteers (aged 35 6 9years; five males, three females) and seven patients (aged47 6 9 years; three males, four females) with clinicaldiagnoses of SpA and ultrasonographically proven Achil-les tendinopathy were studied. The study was approvedby the local ethics commission, and all participants gavewritten informed consent.

The group of volunteers was scanned using the VFA-UTE sequence and SR-UTE sequence in two separate ses-

sions. A reduced number of flip angles (3, 5, 10, 15, 20, and23 or 24�, depending on specific absorption rate (SAR) limi-tations) and TS values (100, 400, 800, and 1200 ms) wereused due to scan time limitations. Total scan times were 24and 50 min for VFA- and SR-UTE sequences, respectively.

The group of patients was scanned using the VFA-UTEsequence only, with optimized flip angles of 4, 19, and 23or 24� (depending on SAR limitations) and TR ¼ 6 ms,giving a total scan time of 9 min. All volunteers andpatients scanned had their ankle positioned above the 4-cm loop coil and comfortably immobilized using MR com-patible foam padding and securing straps attached to theMRI scanner couch. The single slice of the SR-UTEsequence was placed axially through the Achilles tendonapproximately 1 cm above the posterior calcaneus. The3D VFA-UTE field of view was positioned, so that thelowest and highest parts of the slab included the Achillesenthesis and musculo-tendinous junction, respectively.

Four volunteers were scanned using the VFA-UTEsequence with TR ¼ 6 ms and TR ¼ 8 ms to determine anyeffects that TR may have on measured T1 values. Data wereacquired for six flip angles (3, 5, 10, 15, 20, and 24�). To ver-ify simulation results, the same volunteers were scannedusing the optimized flip angle values and TR ¼ 6 and 8 ms.

Image Analysis: Creating Phantom T1 Maps

Data were reconstructed into modulus images. The IR-SEdata were polarity restored to allow for the effect of theinversion pulse on the sign of the longitudinal magnetiza-tion on a voxel-by-voxel basis using a program written inMatlab (MathWorks, Cambridge, United Kingdom; Ref.10). IR-SE signals collected at each TI were fitted to Eqs. 2and 3. SR-UTE signals collected at each TS were fitted toEqs. 2 and 4. At time, t, the signal, S, can be expressed as:

S ¼ S0½1� ð1� cosðaÞÞA� þ s ½2�

A ¼ expð�TI=T1Þ ½3�

A ¼ expð�TS=T1Þ ½4�

where T1 is the longitudinal recovery time, S0 is theequilibrium signal, a is the flip angle of the inversion orsaturation pulse, and s is random noise.

A nonlinear least squares minimization routine using aprogram written in Matlab was used on a voxel-by-voxelbasis with a signal intensity threshold using the longestTI and TS image dataset to avoid fitting to noise and toreduce computational runtime. Start parameters of T1 ¼600 ms, a ¼ 180� for IR-SE and 90� for SR-UTE wereused. The starting value for S0 was based on the voxelintensity corresponding to the voxel from the longest TIor TS dataset. The VFA-UTE data were fitted for T1 andM0 using a model based on Eq. 1 within a programwritten in Matlab, and the same nonlinear least squaresminimization approach as for IR-SE and SR-UTE data. T1

values were obtained using the calculated T1 maps andplacing regions of interest (ROI) within the four agartubes for all sequences.

2 Wright et al.

Image Analysis: In Vivo Data

ROI were drawn around the Achilles on three axial slices ofthe VFA-UTE data; 1 cm above the superior calcanius, 1 cmbelow the musculoskeletal junction and half way betweenusing Analyze (Mayo Clinic, Rochester, MN, USA).

For SR-UTE data, ROI were drawn in each raw inten-sity dataset with mean intensity values fitted for T1

using Eqs. 1 and 3. For the VFA-UTE data, T1 maps werecreated on a voxel-by-voxel basis as for phantom datawith ROI being drawn on the shortest flip angle data setand overlaid onto the chosen slice in the generated T1

map. The highest T1 value of the three ROI was used foreach patient or volunteer for further analysis.

The coefficient of determination (r2) was used to comparebetween the sequences for phantom data. The Mann–Whit-ney U-test was used to compare T1 values from patientsand volunteers calculated for the VFA-UTE sequence.

RESULTS

Simulation and Optimization

Table 1 shows the optimal flip angles for the simulationsat TR ¼ 6 ms and TR ¼ 8 ms, where the difference

between model and mean fitted T1 values were small incomparison to their standard deviations (<15%).The simulation results for the VFA-UTE sequence andTR ¼ 6 ms for combinations of two flip angles are shownin Fig. 1.

Table 1 also shows the different T1 values in fourvolunteer Achilles tendons for different repetition timesin the VFA-UTE sequence and optimized flip angles fortwo (4 and 18�) and three flip angles (4, 19, and 24�)with corresponding T1 values for six flip angles, as usedfor all volunteers.

Phantom Calibration Results

Figure 2 shows T1 values for the four agar phantom tubesusing VFA-UTE and SR-UTE sequences plotted againstthe assumed ‘‘gold standard’’ IR-SE sequence. The T1

values obtained across sequences for each agar phantomtube compare well (r2 ¼ 0.9985 for the VFA-UTEsequence calibration and r2 ¼ 0.9993 for the SR-UTEsequence calibration). However, some scatter wasobserved in the agar tube with the longest T1.

In Vivo Results

The T1 values for volunteer Achilles tendon ROI were725 6 42 ms and 698 6 54 ms, respectively, for the SR-UTE and VFA-UTE sequences, with individual volunteerT1 values shown in Fig. 3. No significant difference[Mann–Whitney U-test] was found between the twosequences for the Achilles tendon measurements.

The mean T1 value calculated using ROI from VFA-UTEdata that gave the highest T1 values was 957 6 173 ms forall patients, compared to 698 6 54 ms for all volunteers(P < 0.01, Mann–Whitney U-test). T1 maps of Achillestendon overlaid on an intensity image (VFA-UTE; 4� flip

Table 1Mean T1 Values Calculated for the Volunteer Subgroup Using the

VFA-UTE Sequence with Two and Three Optimized and Six FlipAngle Values and TRs of 6 and 8 ms

TR (ms) Flip angles (�) Achilles ROI T1 (ms)

6 4, 18 734 6 32

8 4, 18 717 6 406 4, 19, 24 743 6 208 4, 19, 24 723 6 29

6 3, 5, 10, 15, 20, 24 717 6 558 3, 5, 10, 15, 20, 24 709 6 53

FIG. 1. Contour plot of the standard deviation in T1 calculated bysimulating a combination of two flip angles from 1 to 28� for TR ¼6 ms. The lowest standard deviation is marked with an * andstandard deviation errors above 1500 ms have been thresholded

to aid the display of results. [Color figure can be viewed in theonline issue, which is available at wileyonlinelibrary.com.]

FIG. 2. Comparison of SR-UTE and VFA-UTE data to the

assumed ‘‘gold standard’’ IR-SE data obtained in agar phantoms.Error bars indicate standard deviation in T1 of measured ROI.

UTE Sequence Comparison Measuring T1 at 3 T 3

angle) are shown in Fig. 4 for a patient (a) and volunteer(b) with corresponding T2-weighted TSE images.

DISCUSSION

This study has shown that it is possible to measure T1 ofthe human Achilles tendon using UTE imaging techni-ques. The phantom and volunteer results have showngood agreement between SR-UTE and VFA-UTE sequen-ces with mean T1 values being significantly greater in

patients with tendinopathy due to SpA than in asymp-tomatic volunteers.

The longer T1 values demonstrated in patients withAchilles tendinopathy is consistent with previouslyreported reductions in signal on T1-weighted UTEimages (11). This may reflect an increase in water con-tent similar to that previously shown biochemically inmechanical tendinosis of the Achilles tendon (12). Previ-ously published work has suggested that T2 contrast maybe more sensitive than T1 contrast for detection of

FIG. 3. T1 values for the Achilles tendon of volunteers (V) for SR-UTE and VFA-UTE sequences and for patients (P) for optimized VFA-UTE sequence. The error bars indicate the standard deviation of the group mean.

FIG. 4. Patient (a) and asymptomatic volunteer (b) T1 maps overlaid on intensity image. A is Achilles tendon; B is gastrocnemius mus-cle. Corresponding T2-weighted TSE images are shown in (c) and (d).

4 Wright et al.

tendon pathology in SpA (11), but it did not use sequen-ces that were optimized for maximum T1 contrast. Inthis study, direct measurement of T1 relaxation times hasbeen used to maximize image contrast. Furthermore,direct estimates of T1 provide quantitative measurementsthat may reflect tendon disease, with the techniques pre-sented here for quantifying T1 in the Achilles tendonhaving the potential to be extended to other tendonsthroughout the body. Using a 3D sequence such as theVFA-UTE allows the T1 quantitative maps to be refor-matted in any plane, which would be useful for studyingtendons with complex curved anatomy such as themedial or lateral tendon of the ankle and the rotator cuffof the shoulder. Intravenous administration of gadolin-ium based contrast media has also been advocated forthe assessment of tendon disease (6,13). This can be suc-cessfully combined with UTE imaging to sensitivelydetect tendon pathology (7,14), however, contrastenhancement is associated with well-known risks, dis-comfort and cost (15). Unenhanced T1 measurementsmay provide a more acceptable alternative first line ofinvestigation.

Previous estimates of the T1 of the Achilles tendonhave been published at 3 T using UTE sequences incadaveric specimens (4,5). The Achilles tendon T1 valuespublished by Du et al. and Filho et al. were 598 6 37 msand 611 6 15 ms, respectively, which are lower than thevolunteer Achilles tendon T1 values presented here(724 6 45 ms and 694 6 56 ms for SR-UTE and VFA-UTE, respectively). There could be a number of reasonsfor this, including the freeze–thaw cycle that the samplesunderwent prior to imaging, which could have affectedthe tissue MR properties and hydration and temperatureof the samples during imaging. Another differencebetween the cadaveric specimens and volunteers is thatthe gastrocnemius and soleus muscles that attach to theAchilles tendon were cut in the cadaveric specimens,meaning that the Achilles tendon was no longer underload. The specimens used by Filho et al. were further cutinto slices corresponding to slice thickness of the imagesacquired. This could result in a loss of coherence in thetendon collagen structure, affecting the estimated T1.

T1 measurement in the tendon is challenging becauseof the short T2, which precludes the use of many con-ventional strategies and makes the full effect of inversionpulses difficult to realize. Saturation of short T2 speciesis much more robust, making saturation recovery anattractive alternative. However, even at low resolutionsuch measurements typically require long imaging times(30–60 min) and may therefore not be clinically feasible.In severely diseased tendon, the increase in T2 wouldallow measurement of T1 using, for example, conven-tional gradient echo VFA techniques with short TEs.However, in less severe disease, such T1 estimates wouldbe influenced by T2 as well as T1. Future work couldcompare gradient echo and UTE VFA techniques. VFAimaging allows high-resolution 3D T1 maps of the tendonto be achieved in under 10 min using three optimizedflip angles. The combination of two optimized flip anglestend to agree with the relationship described by Deoniet al. (16), where they show the signal from the idealangles is 71 % of the signal from the Ernst angle. The

optimized flip angles calculated here are at 75% of thesignal at the Ernst angle. Saturation-recovery Look-Locker imaging is a potential alternative approach.

A number of alternative quantitative measurements oftendon pathology have been proposed for assessment oftendon disease. T2 or T2* based measurements are sensi-tive to disease and have been applied to the Achilles ten-don (14). However, estimates of T2* obtained by varyingthe time between excitation and acquisition have thedisadvantage of being sensitive to changes due to time-dependent eddy currents, which may limit accuracy (5)with both T2 and T2* being very sensitive to orientationdue to magic angle effects and can increase when the ori-entation is changed from parallel with B0 to the magicangle of 55� (4,17,18). Changes in T2* at 3 T have beenreported from 1.94 6 0.28 ms with the Achilles tendonat 0� to B0 to 15.25 6 2.13 ms at 55� by Du et al. (4).This limits the utility of T2/T2* based measurements forthe assessment of the majority of tendons that are noteasy to align parallel to the magnetic field, such as therotator cuff tendons of the shoulder. In contrast, T1 isrelatively insensitive to magic angle effects in clinicalcontexts (19) with Du et al. (4) reporting a change in T1

from 598 6 37 ms with the Achilles tendon at 0� to B0 to621 6 44 ms at 55� at 3 T.

MRI is widely used in the assessment of the Achillestendon and surrounding tissues in patients with SpA(20,21). Conventional MRI is excellent at detectingabnormalities of the surrounding tissues, for example,retrocalcaneal bursitis, paratenon edema, or calcanealbone marrow edema. Changes visible within the tendonin SpA include thickening with loss of the normal flat orconcave anterior border and increased signal intensityon T1-weighted images with normal or only slightlyincreased signal intensity on T2-weighted images. Suchabnormalities may be detected even when not clinicallysuspected (20,22). However, the sensitivity of tendon sig-nal changes on conventional MRI is limited by the shortT2 of the normal Achilles tendon (1–2 ms; Refs. 2,5,17,and 23), which means that high signal is only apparentafter a substantial increase in the T2 relaxation time,even on T1-weighted images. T1 mapping with UTEimaging has the potential to detect tendon abnormalityearlier by maximizing the sensitivity to subtle changes intendon structure and composition, which affect the lon-gitudinal relaxation rate.

This study has a number of limitations. The SR-UTEsequence used a saturation pulse, which was comparablein length (1 ms) to the T2 of the normal Achilles tendon[�2 ms (5)]. Consequently, saturation may be incompleteand depend to some extent on the T2 of the tendon.Although this is fitted for, it will reduce the sensitivityof the sequence to T1. However, for the 3D VFA-UTEsequence, the radiofrequency excitation pulse was suffi-ciently short (60 ms) compared with the T2 of the Achil-les tendon to assume minimal signal decay during itsapplication. No account was taken of B1 inhomogeneitythat may introduce errors in the T1 estimates. This effectis more important at high field strengths including 3 Tas used here (24). In this study, the Achilles tendon wasinvestigated, which is superficial throughout its course,and B1 field mapping suggested that flip angles were

UTE Sequence Comparison Measuring T1 at 3 T 5

accurate to �5%, corresponding to a 10% error in T1.However, other tendons that are deeper may be moreaffected, and B1 correction may be advantageous. TheVFA-UTE sequence generates isotropic images withlower in-plane resolution compared with 2D techniques.This may limit the assessment of thin structures such asthe fibrocartilage at the enthesis. The 2D saturation UTEimages are contaminated by out-of-slice signal, particu-larly from fat; this could be reduced using fat suppres-sion techniques. Because of the long acquisition time forthe saturation recovery sequence, SR-UTE and VFA-UTEmeasurements were made at different sessions. The num-ber of patients studied was small, and the patient andvolunteer groups were not matched for age or sex. Therepetition time was different between the patient (6 ms)and volunteer (8 ms) group, although this appeared tohave little effect on the T1 estimates (<3%, Table 1) com-pared with the increase in T1 in tendinopathic tendon(37%). ROI selection could also be improved, as in thisstudy, only three slices of the Achilles tendon wereinvestigated in each patient or volunteer, taking thehighest measured T1 of the three ROI. However, thisapproach was used to standardize slice selection andallow at least one slice to contain diseased tendon in thepatient group, as the exact location of SpA within theAchilles tendon was variable.

Future work could include improving the accuracy ofT1 estimates with VFA by B1 mapping. Other tendonsthat do not lie parallel to B0 could also be investigatedsuch as the rotator cuff in the shoulder. This would,however, present a number of challenges, as the rotatorcuff tendons are less superficial and would require alarger field of view while maintaining resolution.Imaging times would be increased, unless accelerationtechniques are implemented such as sparse sampling ofk-space (17). It would be interesting to compare theresults of T1 mapping with both conventional MRI andultrasound. T1 measurements in mechanical or degenera-tive tendinopathy could be compared to those presentedhere in SpA. Longitudinal studies of tendon T1 lookingat disease progression and response to treatment wouldallow the technique to be assessed as a potential bio-marker for tendon disease.

In conclusion, 3D T1 measurements may be performedin the Achilles tendon in under 10 min using an opti-mized VFA-UTE protocol that compares well with thoseobtained over 1 h using 2D saturation recovery. Thesemeasurements show T1 is greater in patients with symp-tomatic SpA and may therefore be useful as a quantita-tive measure of Achilles disease.

REFERENCES

1. Tan AL, Grainger AJ, Tanner SF, Shelley DM, Pease C, Emery P,

McGonagle D. High-resolution magnetic resonance imaging for the

assessment of hand osteoarthritis. Arthritis Rheum 2005;52:

2355–2365.

2. Robson MD, Gatehouse PD, Bydder M, Bydder GM. Magnetic reso-

nance: an introduction to ultrashort TE (UTE) imaging. J Comput

Assist Tomogr 2003;27:825–846.

3. Tyler DJ, Robson MD, Henkelman RM, Young IR, Bydder GM. Mag-

netic resonance imaging with ultrashort TE (UTE) pulse sequences:

technical considerations. J Magn Reson Imaging 2007;25:279–289.

4. Du J, Pak BC, Znamirowski R, Statum S, Takahashi A, Chung CB,

Bydder GM. Magic angle effect in magnetic resonance imaging of the

Achilles tendon and enthesis. Magn Reson Imaging 2009;27:557–564.

5. Filho GH, Du J, Pak BC, Statum S, Znamorowski R, Haghighi P,

Bydder G, Chung CB. Quantitative characterization of the Achilles

tendon in cadaveric specimens: T1 and T2* measurements using

ultrashort-TE MRI at 3 T. AJR Am J Roentgenol 2009;192:

W117–W124.

6. Marshall H, Howarth C, Larkman DJ, Herlihy AH, Oatridge A,

Bydder GM. Contrast-enhanced magic-angle MR imaging of the

Achilles tendon. AJR Am J Roentgenol 2002;179:187–192.

7. Oatridge A, Herlihy A, Thomas RW, Wallace AL, Puri BK, Larkman

DJ, Bydder GM. Magic angle imaging of the achilles tendon in

patients with chronic tendonopathy. Clin Radiol 2003;58:384–388.

8. Marsaglia G, Tsang W. The Ziggurat method for generating random

variables. J Stat Softw 2000;5:1–7.

9. EEC-CRP. IV. Protocols and test objects for the assessment of MRI

equipment. Magn Reson Imaging 1988;6:195–199.

10. Gowland P, Leach M. A simple method for the restoration of signal

polarity in multi-image inversion recovery sequences for measuring

T1. Magn Reson Med 1991;18:224–231.

11. Hodgson RJ, Grainger AJ, O’Connor PJ, Evans R, Coates L, Marzo-

Ortega H, Helliwell P, McGonagle D, Emery P, Robson MD. Imaging

of the Achilles tendon in spondyloarthritis: a comparison of ultra-

sound and conventional, short and ultrashort echo time MRI with

and without intravenous contrast. Eur Radiol 2011;21:1144–1152.

12. de Mos M, van El B, DeGroot J, Jahr H, van Schie HT, van Arkel ER,

Tol H, Heijboer R, van Osch GJ, Verhaar JA. Achilles tendinosis:

changes in biochemical composition and collagen turnover rate. Am

J Sports Med 2007;35:1549–1556.

13. Shalabi A, Kristoffersen-Wiberg M, Papadogiannakis N, Aspelin P,

Movin T. Dynamic contrast-enhanced MR imaging and histopathol-

ogy in chronic achilles tendinosis. A longitudinal MR study of 15

patients. Acta Radiol 2002;43:198–206.

14. Robson MD, Benjamin M, Gishen P, Bydder GM. Magnetic resonance

imaging of the Achilles tendon using ultrashort TE (UTE) pulse

sequences. Clin Radiol 2004;59:727–735.

15. Prince MR, Zhang HL, Roditi GH, Leiner T, Kucharczyk W. Risk fac-

tors for NSF: a literature review. J Magn Reson Imaging 2009;30:

1298–1308.

16. Deoni SC, Peters TM, Rutt BK. Determination of optimal angles for

variable nutation proton magnetic spin–lattice, T1, and spin–spin, T2,

relaxation times measurement. Magn Reson Med 2004;51:194–199.

17. Du J, Chiang AJ, Chung CB, Statum S, Znamirowski R, Takahashi A,

Bydder GM. Orientational analysis of the Achilles tendon and enthe-

sis using an ultrashort echo time spectroscopic imaging sequence.

Magn Reson Imaging 2010;28:178–184.

18. Henkelman RM, Stanisz GJ, Kim JK, Bronskill MJ. Anisotropy of

NMR properties of tissues. Magn Reson Med 1994;32:592–601.

19. Bydder M, Rahal A, Fullerton GD, Bydder GM. The magic angle

effect: a source of artifact, determinant of image contrast, and tech-

nique for imaging. J Magn Reson Imaging 2007;25:290–300.

20. Erdem CZ, Sarikaya S, Erdem LO, Ozdolap S, Gundogdu S. MR

imaging features of foot involvement in ankylosing spondylitis. Eur J

Radiol 2005;53:110–119.

21. Eshed I, Althoff CE, Feist E, Minden K, Schink T, Hamm B, Hermann

KG. Magnetic resonance imaging of hindfoot involvement in patients

with spondyloarthritides: comparison of low-field and high-field

strength units. Eur J Radiol 2008;65:140–147.

22. Erdem CZ, Tekin NS, Sarikaya S, Erdem LO, Gulec S. MR imaging

features of foot involvement in patients with psoriasis. Eur J Radiol

2008;67:521–525.

23. Du J, Takahashi AM, Chung CB. Ultrashort TE spectroscopic imaging

(UTESI): application to the imaging of short T2 relaxation tissues in

the musculoskeletal system. J Magn Reson Imaging 2009;29:412–421.

24. Tardif C, Stikov N, Levesque I, Pike B. Impact of three B1 mapping

techniques on variable flip angle T1 measurements. Proc ISMRM

2011;19:2745.

6 Wright et al.