Although the use of CT has been reported in FAI, evaluation using MRI is currently the main imaging modality of choice after initial radiography [26]. If we accept that this condition does exist and that overall evaluation of the joint (morphologic, labral, and cartilage abnormalities) is important to many clinicians for clinical decision making, we must aim to produce a reliable assessment of these areas on MRI.

We know that reliable MRI assessments of other areas of the body are already achievable using conventional 1.5-T MRI (knee) or 1.5-T MR arthrography (wrist and shoulder). However, the hip presents a relatively unique challenge because of its large, curved, and closely applied articular surfaces and thinner articular cartilage [27]. These features have led to 1.5-T MR arthrography being the workhorse MRI technique for the assessment of hip internal derangement in most institutions [9, 18, 26].

The conspicuity of any small lesion on imaging is affected by the image resolution, signal-to-noise ratio (SNR), and contrast. SNR and contrast are often considered together using the contrast-to-noise (CNR) ratio. SNR and resolution are intimately related because increasing the field gradients improves the resolution at the expense of increased image noise. At 1.5 T, native CNR may be insufficient to reliably show small structures such as a labral tear. Synovial fluid has a long T2 relaxation time compared with the labrum, so high-contrast images may be obtained using long TEs. However, such sequences may be inefficient in terms of SNR not only because there is signal decay during the TE, but also because the T1 relaxation time of synovial fluid is also long; thus, a long TR is also required to maintain the image contrast. Proton density–weighted imaging with an intermediate TE and an intermediate TR allows images with a higher resolution, higher SNR, or both to be obtained, although contrast between the labrum and fluid may be reduced.

To obtain images with improved resolution, improved contrast, and less noise, dilute gadolinium solution may be injected into the joint. Gadolinium lowers the T1 of the fluid so it appears with high signal on T1-weighted images. High-resolution, high-contrast images of the labral tear may therefore be obtained in acceptable times using T1-weighted or proton density fat-suppressed sequences (Figs. 1A, 1B, 1C, and 1D). Such images may also show articular cartilage delamination with high-signal gadolinium-containing fluid tracking deep to the cartilage.

However, unlike arthrographic techniques in other joints, it is difficult to produce reliable distention and bathing of the most relevant structures [27]. Some institutions have tried to overcome this shortcoming by hip manipulation or weighted traction to encourage contrast material to bathe the acetabular and femoral surfaces [28, 29]. A major aid in recent years has been the marked improvements in flexible surface coil technology that, given the hip’s position and the position of significant surrounding soft tissues, have put this area at a further disadvantage compared with other more accessible joints.

The results of studies in the radiology literature support the use of 1.5-T MR arthrography for hip imaging with good accuracy and reproducibility reported for the detection of labral abnormalities [9, 18, 26, 27, 30, 31]. In the MRI studies to date, along with surgical series, there has been no clear acceptance of a single classification for labral abnormalities or of reproducible nomenclature for describing the positioning of acetabular abnormalities. Tear classification has also varied from the complicated (multiple positions with subsections for degenerative cystic findings and so on) to the relatively simple (e.g., normal, basal, or radial intrasubstance) [7, 9, 18, 32, 33]. Acetabular nomenclature used has included clock face, six sectors, eight radial sectors, three sectors (anterior, superolateral, posterior), and two sectors (anterior and posterior) [9, 27, 32, 33]. However, for these different classifications if we accept a transaxial equator through the mid joint, a labral abnormality is most frequently identified in the anterosuperior quadrant above the equator in all types of impingement and an additional posterior labral abnormality seen in pincer impingement [6, 9]. When reporting findings, I generally prefer to initially use a simple description because the classifications used by our hip referrers vary greatly. For labral tears, I describe them as basal (at the junction with bone) or radial (intrasubstance) with positioning according to a quadrant, with tears most commonly located in the anterosuperior quadrant (Figs. 1A, 1B, 1C, 1D, 2A, 2B, 2C, 2D, 3A, 3B, 3C, and 3D).

A pooled analysis of 1.5-T MR arthrographic studies to date estimates a sensitivity of 83% and specificity of 57% for a labral tear [26]. Although not frequently evaluated, interobserver agreement also appears to be good [9]. Most conventional (nonarthrographic) 1.5-T MRI studies have included relatively few patients, but one series of 86 patients reported that MRI had a sensitivity of 97% and specificity of 33% for labral tears [34]. Other conventional 1.5-T studies using high-resolution sequences, a small FOV, and surface coils have reported good accuracy but these published results have not translated into practical imaging at this field strength for most institutions.

Relatively fewer studies have systematically evaluated 1.5-T MRI and MR arthrographic assessments of articular cartilage. Imaging results for articular cartilage show a lower accuracy compared with labral assessment and, when evaluated, poorer interobserver agreement. This limitation of 1.5-T MRI and MR arthrography is probably recognized by most practicing radiologists (Figs. 2A, 2B, 2C, 2D, 3A, 3B, 3C, and 3D).

The anterosuperior acetabular cartilage seems to be the most relevant to surgeons currently when treating patients, with relatively little mention of a femoral cartilage abnormality in the surgical literature [6–8, 21, 24]. Conventional 1.5-T MRI series have reported variable accuracy for assessment of the anterosuperior acetabular cartilage, with two larger series reporting good sensitivity and good observer agreement; however, one study indicated that 1.5-T MRI consistently underestimated the severity [33] and 1.5-T MR arthrographic series have shown a low sensitivity for cartilage abnormalities, particularly delamination [7, 27, 35, 36]. Contrast material or abnormal linear signal identified within or displacing the acetabular cartilage appears to be very specific for cartilage abnormalities, but neither finding is sensitive presumably because delaminated cartilage can still appose the underlying bone and not allow contrast material to dissect underneath [35] (Figs. 2A, 2B, 2C, and 2D). Most series show that using acetabular remodeling, cystic change, or edema as indirect indicators of chondromalacia leads to overcalling of cartilage abnormalities subsequently seen at surgery [6].

If we accept that 1.5-T MR arthrography can evaluate labral abnormalities and femoral morphologic abnormalities, the challenge for 3-T MRI is to improve cartilage abnormality detection and overall characterization [37]. An additional benefit could be to extend accuracy to all areas of the hip without the need for direct arthrography. For patients, this would improve their experience and decrease potential risk. For departments, this would improve patient throughput and reduce costs by removing the need for the arthrographic procedure.

Moving to higher (i.e., 3-T) field strengths improves the SNR, which, in turn, may be exchanged for higher image resolution. The T1 relaxation times of biologic tissues also tend to increase at higher field strengths, so imaging protocols must be carefully optimized for a particular field strength. If this optimization is done, there is usually a net improvement in CNR at 3 T compared with 1.5 T [38].

Although 1.5-T arthrographic images may show delamination of articular cartilage well, 3-T fat-suppressed proton density–weighted or T2-weighted images are probably better for detecting early cartilage abnormalities related to changes in cartilage composition [39].

The potential disadvantages of higher-field-strength imaging are an increase in image artifacts including chemical-shift artifact, which may be countered using fat suppression, and postsurgical metal artifacts. Energy deposition in the tissues is also increased.

Because most abnormalities of the acetabular labrum and cartilage lie in the anterosuperior quadrant, this area may possibly not be well evaluated on conventional axial, sagittal, and coronal sequences at either field strength [9, 11, 40]. Axial oblique (i.e., parallel to the long axis of the femoral neck) sequences are now routinely performed in most institutions to allow more accurate assessment of the alpha angle and cam deformity [9, 18]. However, the reproducibility of this measurement has not been shown to be good in some series and false-negatives can occur if the main cam deformity lies outside this plane [41, 42]. Ideally surgeons would prefer a 3D model of the femoral head to visualize where the cam deformity is most prominent. Partly to address this deficiency, radial MRI sequences have been introduced in some institutions so that the anterosuperior acetabulum lying outside the routine anatomic planes can be sampled and better evaluated [35, 40]. Radial planes are obtained using an axial scout of the femoral head, with planes fanning through 360°, and particular interest in those planes passing through the anterosuperior acetabular quadrant (Fig. 4). Some institutions use a 3D sequence to reconstruct these data, whereas others prefer individual proton density–weighted images. I currently find radial imaging very helpful in defining focal abnormalities not well evaluated on conventional sequences. At our institution, proton density–weighted sequences are performed with eight radial planes in the anterosuperior quadrant (Fig. 4) to try and provide higher-resolution evaluation of cartilage and labral abnormalities (Figs. 5A, 5B, 5C, 5D, 5E, 6A, 6B, and 6C).

In our institution we are evaluating conventional 3-T MRI and 1.5-T MR arthrography of the hip in a number of symptomatic and asymptomatic patient groups. Our aim is to determine whether the accuracy of conventional (nonarthrographic) 3 T is at least equivalent if not superior to 1.5-T MR arthrography for detecting acetabular and cartilage abnormalities.

The 1.5-T MR arthrographic examination (Avanto, Siemens Healthcare) has a scanning time of 25 minutes using a flexible wraparound receive-only multichannel radiofrequency coil (Body Matrix, Siemens Healthcare). With the patient in a supine position, the coil is secured in place over both hips with straps, both hips are positioned in internal rotation, and the patient is advanced into the magnet feet-first.

All of the MR pulse sequences used a small FOV (20 × 20 cm). The following sequences were performed: axial, coronal, and sagittal T1-weighted spin-echo sequence with fat saturation (TR/TE, 633/11; slice thickness, 3 mm; slice gap, 0.6 mm; matrix, 320 × 224; 2 signal averages); T2-weighted turbo spin-echo (TSE) sequence with fat saturation (3630/81; slice thickness, 3 mm; slice gap 0.6 mm; matrix, 384 × 266; 1 signal average); and axial oblique (i.e., parallel to the long axis of the femoral neck) T1-weighted spin-echo sequence (450/11; slice thickness, 3 mm; slice gap, 0.6 mm; matrix, 320 × 224; 2 signal averages).

The scanning time for the 3-T unit (Verio, Siemens Healthcare) is 29 minutes using a small flexible wraparound receive-only radiofrequency coil. The patient is placed in a position identical to that used for 1.5-T MR arthrography.

The following MR pulse sequences were used, all with a small FOV (15 × 15 cm): coronal, sagittal, and axial proton density–weighted TSE sequence with fat saturation (TR/TE, 2000/23; slice thickness, 3 mm; slice gap, 0.3 mm; matrix, 320 × 320; 1 signal average); coronal T2-weighted TSE sequence with fat saturation (5000/65; slice thickness, 3 mm; slice gap, 0.3 mm; matrix, 320 × 320; 1 signal average); coronal T1-weighted TSE sequence (487/10; slice thickness, 3 mm; slice gap, 0.75 mm; matrix, 256 × 256; 2 signal averages); and two proton density–weighted radial TSE sequences (2000/21; slice thickness, 4 mm; slice gap, 2 mm; matrix, 448 × 448; 3 signal averages).

Seven patients from our series (five with surgical correlation [Figs. 1A, 1B, 1C, 1D, 2A, 2B, 2C, 2D, 3A, 3B, 3C, 3D, 5A, 5B, 5C, 5D, 5E, 6A, 6B, and 6C] and two asymptomatic control subjects) [Figs. 7A and 7B]) are shown.