Magnetic resonance histologic correlation in rotator cuff tendons

Authors


Abstract

Purpose:

To relate histologic changes in rotator cuff tendons to the appearance on T1-weighted as well as fat-suppressed T2-weighted and proton density-weighted magnetic resonance imaging (MRI) sequences.

Materials and Methods:

T1-weighted, fat-suppressed T2-weighted and fat-suppressed proton density-weighted sequences of 18 cadaveric shoulders were acquired. The supraspinatus, infraspinatus, and subscapularis tendons were evaluated histologically. Twenty-six abnormalities were found in 23 of 37 tendons. In addition, histologically normal tendon parts (n = 32), including three segments with normal histology but abnormal MR signal, considered to represent magic angle effects, were defined. All regions of interest (ROIs) were evaluated by two musculoskeletal radiologists independently and blinded to histology.

Results:

In the 26 areas with anatomically intact tendons but abnormal histological findings mucoid degeneration (n = 13), chondroid metaplasia (n = 11), fatty infiltration (n = 1), and foreign-body granuloma (n = 1) after tendon suture were found. Compared to normal tendon, mucoid degeneration was hyperintense on T2-weighted fat-suppressed (P = 0.007) and on proton density-weighted fat-suppressed images (P = 0.006). Chondroid metaplasia was hyperintense compared to normal tendon in all sequences (P < 0.05). Mucoid degeneration was hypointense compared to chondroid metaplasia on T2-weighted fat-suppressed images (P = 0.038) and hypointense compared to magic angle artifacts on T1-weighted images (P = 0.046).

Conclusion:

Chondroid metaplasia of rotator cuff tendons appears to be more common than expected. Both mucoid degeneration and chondroid metaplasia may explain increased tendon signal on MR images of the rotator cuff. J. Magn. Reson. Imaging 2010;32:165–172. © 2010 Wiley-Liss, Inc.

MAGNETIC RESONANCE IMAGING (MRI) signal abnormalities of rotator cuff tendons are common. They may be caused by a number of structural changes but also by artifacts. There are few articles comparing histological findings with signal alterations (1–4). Gagey et al (1) evaluated the histology of macroscopically normal supraspinatus tendons of young (14–28 years old) asymptomatic individuals and described fibrillary degeneration, fibrous dystrophy, and eosinophilic transformation of tendon collagen as early degeneration. Kjellin et al (3) reported that areas with increased signal intensity seen on proton density (PD)-weighted images (without further increased signal intensity on T2-weighted images) and an indistinct margin at the articular side of the tendon correspond to eosinophilic, fibrillary, and mucoid degeneration as well as scarring. Increased signal persisting on T2-weighted images was associated with severe degeneration and tendon fiber disruption.

To our knowledge, currently published histological investigations have been limited to the supraspinatus tendon and limited numbers of sequences (PD-weighted and T2-weighted sequences). Thus, the purpose of our study was to relate histologic changes in rotator cuff tendons to their appearance in T1-weighted as well as fat-suppressed T2-weighted and proton density-weighted MR sequences.

MATERIALS AND METHODS

The specimens were obtained from autopsies. At the involved institution the permission to perform autopsies includes the use of body parts for research purposes. Such permission was obtained from relatives.

Eighteen unembalmed cadaveric shoulder specimens were harvested and deep-frozen at −40°C immediately after death. Due to legal reasons, no information regarding the identity, gender, age, or medical history of the donors was available. After MRI, the infraspinatus, supraspinatus, and subscapularis tendons including their bony attachment to the humerus were removed and evaluated macroscopically and histologically. Because the specimens were used for an orthopedic shoulder surgery course before harvesting, some tendons were lost for histological evaluations. Fourteen supraspinatus, 17 infraspinatus, and six subscapularis tendons were evaluated histologically.

MRI

All specimens were allowed to thaw for 24 hours at room temperature prior to MRI. A 1.5-T system (Avanto, Siemens Medical Solutions, Erlangen, Germany) was employed. A four-element shoulder array coil was used. The specimen was placed in the coil similar to the shoulder of a supine patient to achieve a normal anatomic orientation of the tendons to the main magnetic field.

T1-weighted (T1w) spin-echo sequences (repetition time [TR], 539 msec; echo time [TE], 15 msec; field of view [FOV], 14 × 14 cm; section thickness, 3 mm; matrix size, 512 × 512; number of excitations [NEX], 2; bandwidth [BW], ± 31 kHz), fat-suppressed T2-weighted (T2wfs) fast spin-echo sequences (TR, 3040 msec; TE, 71 msec; FOV, 14 × 14 cm; section thickness, 3 mm; matrix, 512 × 512; NEX, 1; echo train length [ETL], 7; BW, ± 31 kHz), and fat-suppressed proton density-weighted (PDwfs) fast spin-echo sequences (TR, 2640 msec; TE, 15 msec; FOV, 14 × 14 cm; section thickness, 3 mm; matrix, 512 × 512; NEX, 2; ETL, 7; BW, ± 28 kHz) were acquired in the transverse, angled sagittal (parallel to the glenoid joint surface), and angled coronal (perpendicular to the glenoid joint surface) planes. All images were stored and evaluated on a PACS.

Tendon Removal and Macroscopic Evaluation

The removal of the tendons was performed by an orthopedic surgeon with specialization in shoulder orthopedics. During this procedure the tendons were evaluated macroscopically by the orthopedic surgeon and a musculoskeletal radiologist in consensus.

In case of partial or complete tendon tears, the size and location with respect to the bony insertion of the tendons were noted.

Histological Evaluation

Immediately after removal and macroscopic evaluation the tendons were placed in buffered 10% formalin and stored in a refrigerator at 5°C. For staining the specimens were washed in tap water and dehydrated using four alcoholic solutions with increasing concentration.

The tendons were embedded in paraffin and cut longitudinally in 2–4-μm sections through the middle of the tendon using a precision saw (Leica SP 1600; Leica Instruments, Nussloch, Germany) and stained with hematoxylin and eosin (H&E).

Histological evaluation was then performed by a musculoskeletal pathologist with 8 years of experience. Evaluation was performed using a high-quality light microscope (Olympus BX41, Center Valley, PA). Mucoid degeneration was diagnosed in the presence of plump tenocytes with chondroid appearance, and discontinuous, disorganized collagen fibers with or without mucoid ground substance. Fatty infiltration was defined as fat tissue between otherwise normal tendon fibers. Lipoid degeneration was defined as fat tissue between degenerated and fragmented tendon fibers. Chondroid metaplasia was defined as fibrous cartilage without communication to the fibrous cartilage that is normally found at the tendon insertions.

The involved pathologist found 26 locations with pathological changes and determined a region of interest (ROI) for each location: eight ROI in seven different supraspinatus tendons, 17 ROI in 14 different infraspinatus tendons, and one ROI in one subscapularis tendon. In addition, 32 ROIs with histologically normal tendon (Fig. 1) substance were determined, three of which were placed in regions with increased MR signal seen in at least one sequence and assumed to represent magic angle effects due to the orientation of the tendon fibers with respect to the main magnetic field, B0. The distance of each ROI from the bony attachment of the tendon was measured using a light microscope (Leica M420, Leica-Microsystems, Glattbrugg, Switzerland) and an attached digital camera (Leica DFC320, Leica-Microsystems). Length was measured digitally using a dedicated software (IM1000, Leica-Microsystems).

Figure 1.

Histologic appearance of a normal tendon in light microscopy at a 20× magnification using H&E staining. Collagen fiber bundles are arranged in a highly organized parallel manner. Small, spindle-shaped nuclei of tenocytes are scattered in the tissue (arrowheads). There are no fat cells and no extracellular matrix is seen. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

To prevent interaction of tendon tears with histological and radiological readout, all ROIs with normal histology were determined at least 5 mm distant to any abnormality seen on histological images. The location of the ROI with pathologic changes was compared to the location of the tendon tears of macroscopic evaluation. All ROI with pathologic changes were at least 5 mm distant to tendon tears.

Positioning of ROI in MR Images

To identify the correct location of the histologically defined ROI in MR images, the measurements from the bony attachment of the tendon to the ROI were used. A curved line through the middle of the tendon beginning at the most proximal end of the tendon insertion was plotted using a DICOM viewing software (OsiriX, v. 3.2.1, OsiriX Foundation, Geneva, Switzerland). Then a circular ROI with a diameter of 3 mm was placed at the site corresponding to the histological abnormalities.

Analysis of MR Images

Two musculoskeletal radiologists with 20 and 10 years of experience analyzed all previously defined ROI independently and blinded to the results of histology. Signal intensity inside the ROI was graded using the signal intensity of bone, fat, muscle, and joint fluid as the reference (Table 1).

Table 1. Grading of Signal Changes in the Tendon Substance
GradeSignal intensity
T1wT2wfs, PDwfs
  1. T1w, T1-weighted sequence; T2wfs, T2-weighted fat-suppressed sequence; PDwfs, proton density-weighted fat-suppressed sequence.

0Hypointense as cortical bone
1Between cortical bone and deltoid muscle
2Isointense compared to deltoid muscle
3Between deltoid muscle and subcutaneous fatBetween deltoid muscle and joint fluid
4Brighter than (subcutaneous) fatBrighter than joint fluid

Statistical Analysis

Weighted kappa values were calculated in order to describe interreader agreement. MR grading of signal intensity and involved area of tendon diameter of the different types of degeneration were compared using a Wilcoxon signed rank test. SPSS (v. 16.0 mac; Chicago, IL) software was used for statistical analysis.

RESULTS

Thirty-seven tendons were evaluated (14 supraspinatus / 17 infraspinatus / 6 subscapularis). Macroscopically there were five partial tendon tears (1/2/2). Five tendons had a transmural tear (4/1/0).

Histologically, mucoid degeneration (Fig. 2) was found in 13 tendons (two supraspinatus, 11 infraspinatus, 0 subscapularis). Eleven tendons (five supraspinatus, five infraspinatus, one subscapularis) had chondroid metaplasia (Figs. 3, 4), and one (supraspinatus) had fatty infiltration (Fig. 5). Mucoid degeneration and chondroid metaplasia were found simultaneously in three infraspinatus tendons. Chondroid metaplasia was never found at the same location as mucoid degeneration but was found directly distal (closer to the tendon insertion) to the location of mucoid degeneration. In one specimen there was a foreign body granuloma with a piece of suture material from previous rotator cuff repair.

Figure 2.

Mucoid degeneration. A: Mucoid degeneration in an infraspinatus tendon in light microscopy at a 20× magnification using H&E staining. Plumb tenocytes (arrows) are characteristic findings of mucoid degeneration. Tendon fibers are not well defined (arrowheads) and the fibers are wavy. Circular ROIs placed in axial MR images of the same infraspinatus tendon. The signal intensity was graded 1 on T1-weighted (B), 1 on T2-weighted fat-suppressed (C), and 1 on proton density-weighted fat-suppressed (D) images. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 3.

Chondroid metaplasia. A: Chondroid metaplasia in an infraspinatus tendon in light microscopy at a 20× magnification using H&E staining. Groups of chondrocytes (arrow) are scattered in newly formed chondroid matrix (arrowheads). Circular ROIs placed in the MR images of the same infraspinatus tendon. The signal intensity was graded 2 on T1-weighted (B), 2 on T2-weighted fat-suppressed (C), and 2 on proton density-weighted fat-suppressed (D) images.

Figure 4.

Chondroid metaplasia. A: Chondroid metaplasia in an infraspinatus tendon in light microscopy at a 20× magnification using H&E staining. Compared to Fig. 3, there is more chondroid matrix (arrowheads) and only few chondrocytes seen (arrows). In the MR images of the same infraspinatus tendon the chondroid metaplasia can be seen as a nicely delineated area in the distal infraspinatus tendon (arrowheads) in T1w (B), T2wfs (C), and PDwfs (D) images. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 5.

Fatty infiltration in a supraspinatus tendon in light microscopy at a 40× magnification using H&E staining. Fat cells (arrow) lying between normal tendon fibers are seen. There is no evidence for tendon degeneration; tendon fibers are sharply defined (arrowheads) and show a normal structure. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Interreader agreement with regard to signal intensity was moderate in T1w (κ 0.56), almost perfect in T2wfs (κ 0.84), and substantial in PDwfs images (κ 0.75). Evaluation based on T1w (κ 0.24) and PDwfs (κ −0.07) images were not successful. The results of the MR evaluation of the different ROIs, separate for the different histologic changes, are presented in Table 2.

Table 2. MR Evaluation of Signal Intensity and Extent of Signal Changes of the Different Histologic and Morphologic Changes
Histologic or morphologic changeSignal intensityInvolved tendon thickness
Median (Range)Median (Range)
 SequenceR1R2R1R2
  1. Signal intensity evaluation based on the grading of signal changes shown in Table 1. Involved tendon thickness: 1, 0-25%; 2, 25-50%; 3, 50-75%; 4, 75-100%. R1, reader 1; R2, reader 2. T1w, T1-weighted sequence; T2wfs, T2-weighted fat-suppressed sequence; PDwfs, proton density-weighted fat-suppressed sequence.

Normal tendon (without magic angle)T1w2 (1-3)2 (1-2)4 (2-4)3 (1-4)
T2wfs1 (0-3)1 (0-3)3 (1-4)3 (1-4)
PDwfs1 (0-2)1 (0-3)4 (1-4)3 (1-4)
Mucoid degenerationT1w1 (1-2)2 (1-2)4 (4-4)4 (2-4)
T2wfs1 (0-3)2 (0-4)4 (1-4)3 (1-4)
PDwfs2 (1-3)2 (1-4)4 (2-4)3 (2-4)
Chondroid metaplasiaT1w2 (1-3)2 (1-3)4 (2-4)3 (2-4)
T2wfs2 (1-3)2 (1-3)3 (1-4)3 (1-4)
PDwfs2 (1-3)2 (1-3)4 (2-4)3 (2-4)
Fatty infiltrationT1w4 (4-4)4 (4-4)4 (4-4)4 (4-4)
T2wfs3 (3-3)3 (3-3)4 (4-4)4 (4-4)
PDwfs3 (3-3)3 (3-3)4 (4-4)4 (4-4)
Magic angle artifactT1w2 (1-2)2 (2-2)4 (4-4)4 (3-4)
T2wfs2 (2-2)2 (1-2)2 (2-3)4 (3-4)
PDwfs2 (2-2)2 (1-2)3 (2-4)4 (4-4)

Compared to normal tendon, mucoid degeneration was characterized by significantly (Wilcoxon signed rank test, P < 0.05) increased signal on T2wfs (P = 0.007) and PDwfs (P = 0.006) images (Table 3). Mucoid degeneration was slightly hypointense compared to magic angle artifacts on T1w images (P = 0.046), and hypointense compared to chondroid metaplasia on T2wfs images (P = 0.038). The significant differences in signal intensities of the different histologic changes are demonstrated in Table 3.

Table 3. Significant (P < 0.05) Differences in Signal Intensity of the Different Histologic Changes (Wilcoxon Signed Rank Test)
 P (2-tailed)
Intensity in T1-weighted sequence 
 Chondroid metaplasia > Normal tendon0.029
 Magic angle artifact > Mucoid degeneration0.046
Intensity in T2-weighted fat-suppressed sequence 
 Chondroid metaplasia > Normal tendon< 0.001
 Chondroid metaplasia > Magic angle artifact< 0.001
 Chondroid metaplasia > Mucoid degeneration0.038
 Mucoid degeneration > Normal tendon0.007
Intensity Proton density-weighted fat-suppressed sequence
 Chondroid metaplasia > Normal tendon0.001
 Mucoid degeneration > Normal tendon0.006

There was vast overlap concerning the spectrum of signal intensity of the different histological patterns (Fig. 6 6). In ROIs with normal tendon histology (n = 29), signal intensity was ≥3 in three ROIs (10%) on T1w images, in four ROIs (14%) on T2wfs images, and in two ROIs (7%) on PDwfs images. The signal intensity of mucoid degeneration (n = 13) was <3 in 13 ROIs (100%) on T1w images, in 10 ROIs (77%) on T2wfs images, and in 10 ROIs (77%) on PDwfs images. In ROIs with chondroid metaplasia (n = 11), signal intensity was <3 in 10 ROIs (91%) on T1w images, in eight ROIs (73%) on T2wfs images, and in seven ROIs (64%) on PDwfs images.

Figure 6.

Signal intensity of the different histologic changes. κ, average signal intensity. The range of signal intensities is visualized by the vertical lines.

Accordingly, a signal intensity grade 1 or 2 in an ROI on T1w/T2wfs/PDwfs images had a sensitivity of 90%/86%/93% and a specificity of 4%/25%/29% for normal tendon substance. Conversely, a signal intensity grade 3 or 4 in an ROI on T1w/T2wfs/PDwfs images had a sensitivity of 4%/25%/29% and a specificity of 90%/86%/93% for tendon pathology.

DISCUSSION

A number of structural changes can alter MR signal of rotator cuff tendons. Normal tendons are characterized by hypointensity in T1w, T2wfs, and PDwfs MR images. T1w sequences can be applied to identify tendinopathy before tendon tears arise. Compared to normal tendon, tendinopathic tendon segments are typically diffusely hyperintense in T1w images. Unless there is a partial tendon tear, these segments do not show any signal abnormality in T2-weighted images (3).

In our study population, mucoid degeneration was the most common abnormality. This finding is in accordance with prior publications (1–5). Rather unexpectedly, the second most common finding was chondroid metaplasia. Chondroid metaplasia has been described in the tibialis posterior tendon (6, 7). There are only few publications that mention chondroid metaplasia in rotator cuff tendons (8–10). Hashimoto et al (9) reported a frequency of 21% for chondroid metaplasia in supraspinatus tendons. In our specimens, chondroid metaplasia was found in 42% of supraspinatus tendons, 23% of infraspinatus tendons, and 17% of subscapularis tendons. The importance of these lesions is largely unknown. Longo et al (11) suggested that they were potential precursors of a tendon tear. Matthews et al (10) evaluated tears in supraspinatus tendons and found chondroid metaplasia and amyloid deposition more often in large tears. They described an inverse proportional relationship between histological changes indicative of repair and inflammation, such as increased fibroblast cellularity, intimal hyperplasia, increased expression of leukocyte and vascular markers, and the size of the tear. However, it remains unclear if chondroid metaplasia favors the development and enlargement of rotator cuff tears or denotes a reactive change caused by rotator cuff tears. In our specimen, all ROIs with chondroid metaplasia and mucoid degeneration were at a distance of at least 5 mm from any rotator cuff tears. Compared to mucoid degeneration, chondroid metaplasia seems to arise more distally (closer to the insertion) in the tendon. However, because we had only three tendons with both degeneration patterns, this could have happened coincidentally. Altogether, we wonder if mucoid degeneration and chondroid metaplasia really are two different entities, or just two different stages (variants) of the same pathologic process. It could be possible that tendon degeneration starts with a cellular reaction of the tenocytes. In this case the tenocytes could transform to fibroblasts and either start producing a scar or undergo chondroid metaplasia, resulting in production of extracellular matrix. Subsequently, chondroid metaplasia degeneration could occur with resultant cell loss from necrosis and transformation to mucoid degeneration. However, we have no proof of this.

Due to the vast overlap concerning the spectrum of signal intensity of the different histological patterns, signal intensities of grade 3 or 4 are not an appropriate criteria to diagnose tendon derangements. Conversely, signal intensities of grade 1 or 2 do not imply normal tendon histology. Thus, we were unable to stratify normal tendon, mucoid degeneration, and chondroid metaplasia based on differences in signal intensity.

The single area of fatty infiltration was hyperintense on T1w images, as expected, and may mimic mucoid degeneration. Unexpectedly, however, this abnormality was also hyperintense on fat-suppressed images. The histological slides of this lesion were reviewed and a localized, diffuse edema between the fat cells was found. The pathogenesis of this edema remains unknown. However, the most likely explanation was an artifact due to freezing and thawing of the specimen. This situation may not be reproducible in vivo. To the best of our knowledge, there is no publication dealing with the influence of freezing and thawing of the specimen on the signal intensity of the tissues. Based on the low cellularity and metabolic activity of tendons, however, the influence of specimen handling on MR characteristics can be expected to be low when compared to other types of tissue.

The prevalence of fatty infiltration and its influence on mechanical properties of tendons are not known.

Signal intensity may also be increased in normal tendons. Magic angle artifacts have been described for short-TE sequences (4, 12, 13). They are most pronounced where the angle between tendon fibers and the B0 magnetic field is 54.7° (14–16). In addition, interdigitation effects have been described at the musculotendinous junction in the supra- and infraspinatus tendons for short-TE sequences (4, 12, 13). One potential explanation for this increased signal is the variable morphology of the musculoskeletal junction and partial volume effects of muscle and tendon tissues. The signal changes are most pronounced when the shoulder is internally rotated (5). Based on a cadaveric study with MR-histological correlation Kjellin et al (3) stated that areas with high signal intensity on proton density-weighted images and normal signal on T2-weighted (T2w) images with blurred borders related to mucoid degeneration. Conversely, high signal on T2w images most probably indicated severe degeneration and substance defects. If signal intensity on T2-weighted images is higher than adjacent muscle tissue it is more likely to be a tear than a magic angle artifact (17). Our results indicate that similar rules are valid for the more commonly used fat-suppressed spin-echo sequences.

Interreader agreement was best for T2wfs and PDwfs images. In the evaluation of the extent of involved tendon thickness T2wfs images were superior to T1w or PDwfs images.

Study limitations relate to the relatively small number of specimen, which precluded use of a more detailed grading scale for signal intensities. Small differences with regard to sequences and the various types of tendon abnormalities may be underestimated using a grading scale with only four signal intensity grades. The lack of information about age, gender, or medical history of the donors limits the ability to define the symptomatic or asymptomatic aspect of the tendon.

In conclusion, chondroid metaplasia of rotator cuff tendons appears to be more common than expected. Both mucoid degeneration and chondroid metaplasia may explain increased tendon signal on MR images of the rotator cuff.

Ancillary