Rotator Cuff Injury MRI

Shoulder pain is a common complaint of patients during physician visits, and it can be due to a variety of causes. The major cause of shoulder pain in patients older than 40 years is rotator cuff impingement and tears. It is estimated that 20-50% of those 60 years or older have a known rotator cuff tear (RCT), and up to 65% of individuals older than 70 years have an asymptomatic RCT. ^[1]

The number of shoulder arthroscopies is steadily increasing to treat glenohumeral joint disorders, among which RCT is the most common. The prevalence of this condition ranges from 13 to 37% in the general population without taking into account the number of asymptomatic patients. ^[2]

With the development of new arthroscopic techniques for treating rotator cuff disorders, magnetic resonance imaging (MRI) has played an increasingly important role as a noninvasive test for determining which patients may benefit from surgery. ^{[3, 4, 5, 6]}

Rotator cuff pathology is commonly encountered and can manifest as tendinopathy or tearing. Magnetic resonance imaging (MRI) and ultrasonography (US) offer similar sensitivity and specificity for evaluation of the native rotator cuff; the chosen imaging modality may vary, depending on local practice and accessibility. Magnetic resonance arthrography (MRA) is frequently used in the postoperative setting as a problem-solving tool. ^[7]

Gilat et al found that ultrasonography had a sensitivity of 80.8% and a specificity of 100% for the detection of recurrent rotator cuff re-tears in patients with shoulder pain after rotator cuff repair (RCR). They performed a retrospective analysis of 39 patients with shoulder pain after arthroscopic RCR who had undergone US, followed by revision arthroscopy, and a rotator cuff re-tear was indicated by US in 21 patients (54%) and by revision arthroscopy in 26 patients (67%). ^[8]

Repair of RCTs has been shown to improve functional outcomes in full-thickness tears. High re-tear rates have been reported, but this does not correlate with poorer outcomes. Paul and associates explored the association between functional outcome scores and structural integrity of the rotator cuff using MRI-based grading. They found no association between postoperative structural integrity of the repaired tendon and functional outcome scores and noted that early repair of the rotator cuff results in better functional outcomes and shoulder strength. ^[9]

(See the images below.)

Lambert et al found the positive predictive value of 3.0T MRI to be 100% for detection of rotator cuff tendon tears requiring surgery. In this prospective, follow-up study of 48 patients, when arthroscopy was performed based on MRI findings, there was no change in surgical management from that determined by MRI. ^[3]

Yoo et al found that preoperative MRI variables may help predict incomplete arthroscopic repair of large to massive RCTs. On preoperative MRIs of RCTs, these authors found that fatty degeneration index (FDI) values greater than 3 on sagittal oblique sections of the supraspinatus and values greater than 2 on sagittal oblique sections of the infraspinatus, with greater than 31 mm in coronal oblique tear distance (COTD) and 32 mm in sagittal oblique tear distance (SOTD), can help predict incomplete arthroscopic repair of the torn tendon. ^{[10, 11]}

According to the American College of Radiology (ACR), the initial imaging modality for traumatic or atraumatic shoulder pain should be radiography, and the next imaging choice should be guided by the clinical scenario and by findings on the plain films. ^{[12, 13, 14]}

Conventional MRI

Conventional MRI with T2-weighted images in the oblique coronal and oblique sagittal planes is the preferred technique for imaging the rotator cuff. Most authors have found that fat-suppressed, fast spin-echo (FSE), T2-weighted images are most accurate for detecting RCTs; sensitivity of 84-100% and specificity of at least 77-97% for full-thickness tears can be expected with this pulse sequence. ^{[15, 16, 17, 18, 19, 20, 21, 22]}

Although most RCTs can be seen on oblique coronal images, Patten et al reported that oblique sagittal images provide approximately 10% improvement in accuracy for detecting RCTs, although this finding was not statistically significant. ^[23] These authors reported that oblique sagittal images are especially helpful for identifying tears involving the anterior edge of the supraspinatus.

Magnetic resonance arthrography

MRI is the first-line imaging modality for assessing joints because of its superior soft tissue contrast capability. CT arthrography (CTA) is used when there is a contraindication to magnetic resonance arthrography (MRA), such as for a patient with incompatible vascular clips, claustrophobia, or a pacemaker, or when MRA is unavailable. ^[12]

Some clinicians prefer to perform direct or indirect MRA for imaging the rotator cuff. The advantage of direct MRA relative to MRI is that it distends the joint, thus forcing contrast agent into a small defect. T1-weighted images, which are faster to acquire and have a superior signal-to-noise ratio, can be used instead of T2-weighted images. The disadvantages of direct MRA are that it is mildly invasive and that it may require imaging guidance to place a needle into the glenohumeral joint capsule. In addition, bursal surface partial-thickness tears are not directly opacified. ^{[4, 11, 17, 22, 24, 25]}

Several authors have reported that direct MRA is close to 100% sensitive and specific for full-thickness and articular surface partial-thickness RCTs. ^[26] A full-thickness tear will demonstrate the gadolinium contrast solution extending first through a defect in the cuff and then into the subacromial-subdeltoid bursa. Articular surface partial-thickness tears show focal extension of the contrast solution into the substance of the tendon.

(See the image below.)

When performing direct MRA, it is important to use fat suppression to decrease the signal intensity of the peribursal fat plane around the subacromial-subdeltoid bursa; without fat suppression, the fat plane can mimic the contrast agent, leading to false interpretation of an RCT.

In a meta-analysis of studies on MRI, MRA, and ultrasonography for RCTs, de Jesus et al found MRA to be more sensitive and specific than either MRI or US for diagnosing both full-thickness and partial-thickness tears. MRI and US showed no significant differences in sensitivity or specificity for full- or partial-thickness tears. ^[4]

Indirect MRA requires only an intravenous (IV) injection, but this modality has a disadvantage in that it does not distend the joint. As in direct MRA, short scanning time T1-weighted images can be used instead of T2-weighted images. Several authors have shown that compared with conventional MRIs of the rotator cuff, RCTs are better characterized on indirect MRA and there is better correlation with surgical findings. One study reported that 2 radiologists improved their accuracy for detecting RCTs from 67% and 62% with conventional MRI to 92% and 96%, respectively, with indirect MRA. ^[27] Again, use of fat suppression is important, but exercising the joint does not appear to improve accuracy.

Despite these studies, MRA has not been as widely accepted for evaluating the rotator cuff as it has been for imaging the glenoid labrum. Direct MRA does improve the depiction of posterior articular surface partial-thickness tears that are observed in overhead-throwing athletes, particularly if the shoulder is scanned in abduction and external rotation. However, most authors have found that fat-suppressed, FSE, T2-weighted images obtained with a quality shoulder coil are fairly accurate for most RCTs, and that conventional MRI is adequate for routine imaging of the rotator cuff.

Conventional arthrography was the traditional technique for detecting RCTs. However, arthrography itself does not demonstrate bursal-sided, partial-thickness tears, and it may be difficult at times to determine the size of a tear using this modality. With improvements in CT scanners, oblique coronal reformatted CTA can provide excellent images of the rotator cuff in patients who are unable to undergo MRI.

Limitations of techniques

MRI is contraindicated in patients who have a cardiac pacemaker, ferromagnetic foreign bodies (particularly in the orbit), and some cochlear implants. Some patients are extremely claustrophobic in high-field-strength MRI scanners, although many of these patients can be scanned in open MRI scanners after administration of a mild sedative.

MRA is mildly invasive, and imaging is usually necessary to correctly position the arthrogram needle within the joint capsule. Fluoroscopy is the most common way to provide imaging guidance, but needle placement also can be performed under CT scanning, with ultrasound, or within the MRI scanner. Conventional arthrography is mildly invasive and has the limitation of not being a tomographic technique. Although fluoroscopy has been used traditionally, the involvement of ionizing radiation may represent a limitation for use of this modality. US has emerged as a feasible, radiation-free modality that can be used to inject joints, with accuracy comparable to fluoroscopy. ^[28]  

Gadolinium-based contrast agents have been linked to the development of nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD). This disease has occurred in patients with moderate to end-stage renal disease after they were given a gadolinium-based contrast agent to enhance MRI or MRA scans. NSF/NFD is a debilitating and sometimes fatal disease. Characteristics include red or dark patches on the skin; burning, itching, swelling, hardening, and tightening of the skin; yellow spots on the whites of the eyes; joint stiffness with trouble moving or straightening the arms, hands, legs, or feet; pain deep in the hip bones or ribs; and muscle weakness. 

An FDA warning was added to labels that gadolinium may be retained in various organs, including the brain, after use of these agents during MRI. Linear gadolinium-based contrast agents (GBCAs) present greater risk than macrocyclic agents. The warning notes greater risk in specific patient populations, including children and pregnant women. ^[29]  In addition, the European Medicines Agency (EMA) followed recommendations from its Pharmacovigilance Risk Assessment Committee and Committee for Medicinal Products for Human Use to restrict use of the IV linear agents gadoxetic acid and gadobenic acid, as well as gadopentetic acid given intra-articularly. The EMA has recommended suspension of all other linear products. ^[29]

Magic-angle effect

The histology of the rotator cuff contributes to one of the difficulties of rotator cuff MRI interpretation—the magic-angle effect, or angular anisotropy. This effect is an MRI artifact in which normally low-signal structures that are made of organized collagen fibers appear with higher signal intensity on images obtained with a short echo time (TE). This artifact occurs when the long axes of collagen fibers are oriented at 55° to the main magnetic field.

In most high-field MRI scanners, the main magnetic field is oriented along the direction of the bore (the central tunnel where the patient lies). Well-organized collagen fibers in the outer portions of the rotator cuff are organized longitudinally; therefore, these normally low-signal fibers have increased signal intensity on short-TE images as the fibers curve and become oriented at the magic angle.

Unfortunately, this effect occurs in the region of the critical zone where RCTs and degenerative tendinopathy are prevalent. However, the magic angle’s high signal intensity diminishes with increasing TE; thus, it is not usually a problem on the fat-suppressed, FSE, T2-weighted MRIs most radiologists currently use to image the rotator cuff.

Full-thickness tears

For many orthopedic surgeons, the main role of shoulder MRI is to detect a full-thickness rotator cuff tear (RCT). The most common appearance of a full-thickness tear is high signal intensity on a T2-weighted image that extends from the articular surface of the rotator cuff to the subacromial-subdeltoid bursa.

(See the image below.)

Rafii et al reported that high signal was observed in approximately 90% of full-thickness tears proven at surgery. ^[30] In chronic RCTs in which the shoulder joint has little or no effusion, the humeral head may be high riding, such that not much high signal is seen at the tear site.

(See the image below.)

Some patients may develop fibrous thickening of the subacromial-subdeltoid bursa, which can mimic an intact tendon in the absence of an effusion; therefore, it is important to trace a low-signal structure as it passes over the humeral head. Rotator cuff fibers will end at their insertion on the greater tuberosity, whereas fibrous thickening of the bursa will continue deep to the deltoid muscle below the greater tuberosity.

In addition, acute RCTs can hemorrhage at the tear site, with the blood mimicking some intact fibers. It is important to distinguish the smoothly curving, low-signal surfaces of the rotator cuff from the disorganized low-signal surfaces of fibrin and other blood products.

Most small full-thickness tears arise in the anterior aspect of the supraspinatus tendon in the critical zone.

(See the images below.)

Localizing a small full-thickness tear to the rotator cuff crescent may be helpful for the shoulder surgeon, who may then decide to only debride, but not repair, the cuff defect. Although RCTs often begin in the critical zone, resorption of the tendon stump at the greater tuberosity may occur if chronic full-thickness tears are left untreated. Full-thickness avulsion tears of the tendon away from the greater tuberosity are less common. Massive tears often extend posteriorly to involve the infraspinatus tendon or extend anteriorly to tear the anterior interval and the subscapularis tendon.

If a full-thickness tear is observed, it is important to document whether or not the entire anterior-to-posterior width of the supraspinatus tendon is involved. In RCTs that involve the entire tendon, the tendon edge can retract medial to the glenoid, where it becomes extremely difficult to grasp and to reattach to the greater tuberosity.

Long-standing RCTs can result in muscle atrophy and fatty degeneration that may prevent successful repair. It is important to expeditiously obtain imaging studies for patients who have a possible acute full-thickness, complete-width, supraspinatus tendon tear. If an acute complete supraspinatus tendon tear is identified, surgery is often scheduled within the next several days, so the tendon can be repaired before retraction or atrophy occurs.

Partial-thickness tears

Partial-thickness tears can be classified as articular, bursal, or intratendinous. Intratendinous tears may be the cause of shoulder pain, but they are not observed at routine arthroscopy and are rarely treated surgically. Articular surface partial-thickness tears are more common than bursal surface tears (at an approximately 3:1 incidence rate). ^[31] Many patients with a bursal surface tear also have an articular surface tear.

The accuracy of MRI for partial-thickness tears is less than that for full-thickness tears. Although some authors have reported sensitivity greater than 0.90 for partial-thickness tears, others have reported sensitivities as low as 0.17-0.56. ^{[30, 32, 33, 34, 35]}

Reinus et al were unable to correctly identify the side of the affected rotator cuff (articular vs bursal) in 50% of patients with a partial-thickness tear. ^[36] One reason for this low accuracy is that a high-signal defect on T2-weighted images is a less common finding in partial-thickness tears than in full-thickness tears; in a study by Rafii et al, this high-signal defect was seen in only 7 of 16 cases of partial-thickness tears. ^[30]

Partial-thickness tears often appear on MRI as only an intermediate signal, isointense to muscle, which disrupts the normal low-signal surface of the rotator cuff.

Absence of fluid in an RCT on MRI may be caused by the presence of a poor-quality scar or granulation tissue within the defect, and this can be difficult to distinguish from tendon degeneration or a healed RCT. Partial-thickness tears may have smooth margins that taper gradually, so that the rotator cuff appears to be only somewhat thinned.

(See the image below.)

Although most partial-thickness tears occur in the critical zone of the supraspinatus tendon, some RCTs occur in less common locations. Among younger patients, a small articular surface avulsion-type partial-thickness tear can occur adjacent to the greater tuberosity; this is referred to as a "rim-rent tear."

(See the image below.)

Tears isolated to the infraspinatus tendon occur in 1-7% of patients with RCTs, but these tears are more common among athletes who perform overhead activities. ^[37] MRA in which the patient is positioned with the arm in abduction and external rotation is the best technique for identifying these infraspinatus articular surface partial-thickness tears, which often are associated with adjacent glenoid labral fraying.

Although the most important MRI criterion of a partial-thickness tear is the presence of an increased signal that disrupts the normally low-signal surface of the rotator cuff, some authors have described secondary signs that may be helpful in improving the accuracy of MRI. Sanders et al demonstrated that an intramuscular cyst, typically in the supraspinatus muscle, is always associated with articular surface involvement by a tear. ^[38] These authors suggested that when such cysts are present, associated rotator cuff pathology should be investigated.

Inferiorly directed acromioclavicular joint osteophytes, a hooked anterior acromion, and an os acromiale have all been associated with a higher incidence of RCTs; therefore, these findings should prompt careful evaluation of the rotator cuff. ^[39] Subacromial-subdeltoid fluid is common in full-thickness tears, but a small amount can be observed in patients without a bursal surface tear; thus, the presence of this fluid is not an accurate secondary sign of a bursal surface partial-thickness tear.

Degree of confidence

Studies investigating the use of fat-suppressed, FSE imaging have reported sensitivity of 84-100% and specificity of 77-97% for full-thickness tears ^{[32, 33, 40, 41]} ; however, accuracy for partial-thickness tears is lower. MRA may be helpful for better demonstrating articular surface partial-thickness tears. Angling oblique coronal or oblique sagittal images to the rotator cuff surface at the suspected tear site can improve the accuracy of conventional MRI.

False positives/negatives

Three other abnormalities of the rotator cuff can mimic an RCT: degeneration, tendinopathy, and cuff strain. Rotator cuff degeneration is common among older individuals and appears as an ill-defined area of increased signal on T2-weighted MRIs within the substance of the cuff. All rotator cuffs undergo age-related degeneration in which the normally compact and well-organized collagen fibers are replaced by intermediate-signal myxoid and eosinophilic material.

As aging progresses and the rotator cuff is put under repeated stress, small fissures can develop within the cuff substance and appear as thin areas of fluid on MRI. If MRI contrast and brightness are set too high (ie, windowed too tightly), these fissures occasionally bloom and may appear as a tear that extends to the surface of the cuff.

(See the image below.)

Tendinopathy, occasionally incorrectly termed tendinitis, is a related intratendinous process that is histologically similar to rotator cuff degeneration. Although the term tendinopathy is occasionally used interchangeably with age-related cuff degeneration, some clinicians reserve the term for younger symptomatic patients.

As with patellar "tendinitis," tendinopathy is not truly an inflammatory process because there is no edema, vascular invasion, or acute inflammatory cells. Instead, what occurs pathologically is severe mucoid and eosinophilic degeneration with intratendinous clefts, often causing focal tendon swelling and, occasionally, surface fibrillation. If windowed incorrectly during imaging, tendinopathy can appear to extend to involve the surface of the rotator cuff.

(See the images below.)

Rotator cuff strain after acute trauma has been described as another potential cause of increased intratendinous signal on MRI. This typically occurs in younger patients (< 35 yr) who have an associated bone bruise and focal increased signal intensity in the posterior aspect of the supraspinatus tendon, as distinguished from cuff degeneration, which involves a larger area that is centered in the anterior critical zone. Patients with presumed rotator cuff strain as demonstrated on MRI are less likely to require surgery than older patients who develop shoulder pain after acute trauma.

Yu et al performed a retrospective analysis to assess the correlation between calcific tendinitis and RCTs using MRI of shoulder joints. These authors found no significant correlation between the calcification site and RCTs in the calcific tendinitis group and noted that only 3.7% of patients showed calcification and a tear in the exact location of the same tendon (P>0.05, r = 0.03). They concluded that compared to patients with shoulder pain without calcific tendinitis, no increase in risk of RCT was evident on MRI in those with rotator cuff calcific tendinitis; therefore, calcific tendinitis and RCT may have different pathologic causes, and there is no significant correlation between them. ^[42]

MRI is generally considered the modality of choice for assessment of rotator cuff pathology, including tendinosis. Assessment of severity is currently subjective and is based primarily on tendon morphology and signal intensity. Pow and colleagues explored a novel objective means of evaluating rotator cuff tendinosis severity and reported correlation of their findings with subjective assessment of severity. They concluded that further research is required to assess the utility of such objective measurements across varying imaging protocols with awareness of inherent limitations of quantifying signal intensity on MRI. ^[43]

In summary, fat-suppressed, FSE, T2-weighted images obtained with a quality shoulder coil are accurate for diagnosing RCTs. False-negative full-thickness tears typically occur when the patient does not have an effusion and when the subdeltoid bursal capsule is thickened. False-negative partial-thickness tears are fairly common, especially for tears that are not very deep. Failure to diagnose partial-thickness tears can be minimized by radiologists carefully inspecting the low-signal surfaces of the rotator cuff and noting whether the low-signal surface layers are disrupted, as well as by use of both intra-articular and IV gadolinium to enhance the conspicuity of these lesions.

Three-dimensional MRI (dynamic 3D MRI) enables noninvasive monitoring of the kinematics of the shoulder complex during slow active arm elevation. The acquisition process of the dynamic MRI technique was found to be fast enough to carry out multiple scans while the patient abducted the shoulder in a continuous motion. ^[44]  

In a study of 20 shoulders (4 shoulders with massive rotator cuff tears, 5 shoulders with an isolated full-thickness supraspinatus tear, 5 shoulders with tendinopathy, and 6 normal shoulders), a real-time 3D image series was conducted to measure changes in the width of subacromial space, superior-inferior translation, and anterior-posterior translation of the humeral head relative to the glenoid during active abduction. These measures were investigated for consistency with rotator cuff disease classifications from standard MRI. Use of dynamic MRI enabled a novel measure the investigators termed "looseness," which they defined as the translation of the humeral head on the glenoid during an abduction cycle. Looseness was better able to differentiate different forms of rotator cuff disease than a simple static measure of relative glenohumeral position. ^[45]

Wallenberg et al evaluated MRI-based 3D volumetric assessment of fatty infiltration and muscle atrophy in RCTs. They noted that the Goutallier and Warner Classification systems are useful in determining rotator cuff reparability. Upon performing image analysis of 109 patients who underwent shoulder MRI, they found that parasagittal imaging location did not significantly influence the Goutallier score, but assessment of muscle atrophy using the Warner score led to the perception of less atrophy medially regardless of the magnitude of tendon retraction. ^[46]

Kim and colleagues noted that 3D MRI offers more intuitive visualization and can better facilitate treatment planning than 2D MRI. By using a convolutional neural network to visualize 3D models of related anatomic structures, the investigators found that automated segmentation using 3D nnU-Net, an automatic segmentation method based on deep learning, produced acceptable accuracy and reproducibility. ^[47]