Adhesive capsulitis is among the most common diagnoses in patients seen in orthopedic shoulder and sports medicine clinics [1–5],but is much less frequently reported at MRI. MRI has been proven sensitive and specific for shoulder lesions including rotator cuff tendinopathy, glenoid labrum tears, and arthropathies, but its value in confirming a diagnosis of adhesive capsulitis has not been clearly established. Direct and indirectMR arthrography has been shown to accurately diagnose adhesive capsulitis or frozen shoulder [17–20], but adhesive capsulitis is most prevalent in the 45 to 60-year-old age group, a population forwhich direct or indirectMR arthrography is rarely ordered. Effective nonoperative treatment regimens for adhesive capsulitis have been established, and delayed diagnosis can be a cause of significant morbidity and protracted syndromes [9,12,16,28]. Given the prevalence of concomitant pathologies such as impingement, rotator cuff tendinopathy, and degenerative arthropathies in this population,routine noncontrast shoulder MRI examinations are frequently ordered in patients with clinical adhesive capsulitis. While some MRI findings of adhesive capsulitis have been described previously, we sought to establish a specific algorithm for confidently confirming the presence of adhesive capsulitis with routine, noncontrast MRI.Rotator cuff interval findings in adhesive capsulitis including thickening of the coracohumeral ligament itself and fibrosis or synovitis in the rotator intervalfat have been described in the orthopedics literature based on arthroscopy and open surgical evaluation [11–16]. The anatomy of these structures is thought to play an important role in the restriction of external rotation of the shoulder seen in the setting of adhesive capsulitis [10, 14,29]. Our study shows a statistically significant difference in coracohumeral ligament thickness between study group subjects with clinical adhesive capsulitis compared to control subjects,with high sensitivity and reasonable specificity. Observation of rotator interval infiltration of the subcoracoid fat in conjunction with coracohumeral ligament thickeningmarginally improved specificity for detection of adhesive capsulitis; however, there was not a statistically significant difference in sensitivities or specificities based on one versus two criteria assessment. GondimTeixeira et al. previously reported low sensitivities for detection of adhesive capsulitis based on coracohumeral ligament thickness N4 mm[26]. However, the cut off coracohumeral thickness may have significantly traded off sensitivity for specificity in this series.Contrary to the findings at the rotator interval, imaging findings associated with adhesive capsulitis at the axillary recess have been debated in the literature. Several studies have described the MRI findings of thickening and edema of the axillary recess or inferior glenohumeral ligament in adhesive capsulitis [20,22,25,26], while other studies have not found similar findings [17,19]. Song et al. and Gondim et al. described high sensitivity and specificity for adhesive capsulitis based on thickening of the axillary recess and increased T2 signal of the inferior glenohumeral ligament [20,26], respectively, while Sofka et al. demonstrated similar thickening and signal of the axillary recess in only one of the four clinical stages of adhesive capsulitis [25]. In our study, a constellation of observations including coracohumeral ligament thickening, rotator interval infiltration, and thickening/edema of the axillary recess yielded high specificity and a statistically significant difference in sensitivity and specificity compared to using one or two criteria alone. The axillary recess was considered thickened if it measured N2 mm, noting variability ofmeasurements of the axillary recess reported in the literature [19,22,24,25]. As this measurement can be difficult because patients with adhesive capsulitis often do not externally rotate on the scanner, axillary recess pericapsular edema was also included in our analysis. The increased specificity with inclusion of the axillary recess findings suggests that axillary recess involvementmay be seen as synovitis and fibrosis at the rotator interval evolves. Routine noncontrast MRI shoulder examinations provide multiple imaging planes and both fat suppressed and non-fat suppressed sequences, ideal for an algorithmic approach in the assessment for adhesive capsulitis. As such, the use of both specificMRI findings and constellations ofMRI findings is practical to maximize accurate diagnosis of adhesive capsulitis. Based on the results of our study, clinical criteria for adhesive capsulitis can be used in conjunction with a practical 1, 2, or 3 criteria noncontrast MRI system, guided by the level of clinical suspicion. In the setting of high clinical suspicion for adhesive capsulitis, the isolated finding of coracohumeral ligament thickening N2 mm yields a strong sensitivity for a confirmatory diagnosis. On the contrary, coracohumeral ligament thickness equal to or b2mmmay help to exclude adhesive capsulitis. In the setting ofmoderate clinical suspicion for adhesive capsulitis, the addition of infiltration of the rotator interval fat may be useful to improve confidence of diagnosis.In the setting of atypical or low initial clinical suspicion for adhesive capsulitis, the constellation of findings of axillary recess thickening/ edema in conjunction with rotator cuff interval findings of coracohumeral ligament thickening and infiltration of the rotator interval fat strongly suggests a diagnosis of adhesive capsulitis with high specificity and may raise suspicion for adhesive capsulitis. Implementing this system at our institution for 500 consecutive routine MRI shoulder examination has yielded a prevalence of adhesive capsulitis in ourMRI reports of 6.8%, similar to the estimated prevalence of adhesive capsulitis in orthopedic shoulder clinic of 5–6% [8,19].There are several limitations to our study.There was no arthroscopic or surgical pathology correlation in our study. However, as adhesive capsulitis is generally not treated with surgery,a reference standard of clinical adhesive capsulitis based upon subspecialty physical examination was used, which has been accepted in recent orthopedic literature. Given that current treatment for adhesive capsulitis is largely directed toward physical therapy and percutaneous steroid/anesthetic injections, histological specimens were not available for analysis. Additionally, given the retrospective design of our study, information regarding the clinical stage of adhesive capsulitis at the time of imaging was not available to review.

5. Conclusion

 In conclusion, adhesive capsulitis can and should be accurately and consistently diagnosed on routine noncontrast shoulder MRI in conjunction with appropriate clinical criteria. The finding of a thickened coracohumeral ligament shows strong sensitivity for adhesive capsulitis while the constellation of coracohumeral ligament thickening, rotator interval infiltration of the subcoracoid fat, and axillary recess thickening/edema yields great specificity for adhesive capsulitis.

粘连性关节囊炎是骨科肩关节和运动医学诊所最常见的诊断之一[1-5]，但在MRI上报告的频率要低得多。MRI已被证明对肩关节病变（包括肩袖肌腱病、盂唇撕裂和关节病）敏感和特异，但其在确认粘连性囊炎诊断中的价值尚未明确确立。直接和间接MR关节造影已被证明能准确诊断粘连性囊炎或冻结肩[17-20]，但粘连性荚膜炎最常见于45-60岁年龄组，这一人群很少要求直接或间接MR膝关节造影。粘连性囊炎的有效非手术治疗方案已经建立，延迟诊断可能是导致严重发病率和长期综合征的原因[9,12,16,28]。考虑到该人群中合并病变如撞击、肩袖肌腱病和退行性关节病的普遍性，临床粘着性关节囊炎患者经常需要常规非对照肩MRI检查。虽然之前已经描述了粘连性囊炎的一些MRI发现，但我们试图建立一种特定的算法，用常规非对比MRI自信地确认粘连性囊病的存在。

基于关节镜和开放手术评估的骨科文献中描述了粘连性关节囊炎的肩袖间隔发现，包括喙肱韧带本身增厚和肩间脂肪纤维化或滑膜炎[11-16]。这些结构的解剖结构被认为在限制粘连性囊炎中可见的肩部外部旋转方面起着重要作用[10，14，29]。我们的研究表明，与对照组受试者相比，临床粘连性囊炎受试者的喙肱韧带厚度存在统计学显著差异，具有较高的敏感性和合理的特异性。观察旋转肌间隙、椎下脂肪浸润以及喙肱韧带增厚，可提高检测粘连性囊炎的特异性；然而，基于单标准和双标准评估的敏感性或特异性没有统计学显著差异。

GondimTeixeira等人此前报道，基于喙数韧带厚度>4 mm的粘连性囊炎检测灵敏度较低[26]。

然而，在这一系列中，切掉的喙肱厚度可能在敏感性和特异性之间有显著的折衷。与旋转间隔期的发现相反，与腋窝隐窝粘连性囊炎相关的影像学发现在文献中一直存在争议。一些研究描述了粘连性囊炎患者腋窝隐窝或下盂肱韧带增厚和水肿的MRI表现[20,22,25,26]，而其他研究未发现类似的表现[17,19]。Song等人和Gondim等人分别描述了基于腋窝隐窝增厚和下盂肱韧带T2信号增加的粘附性囊炎的高灵敏度和特异性[20,26]，而Sofka等人仅在粘附性囊病的四个临床阶段中的一个阶段显示了类似的腋窝隐匿增厚和信号[25]。在我们的研究中，包括喙肱韧带增厚、旋转肌间隙浸润和腋窝隐窝增厚/水肿在内的一系列观察结果产生了高特异性，与单独使用一个或两个标准相比，敏感性和特异性具有统计学显著差异。

如果腋窝测量值>2mm，则认为腋窝增厚，注意文献[19,22,24,25]中所述腋窝测量的可变性。

由于粘着性囊炎患者通常不会在扫描仪上向外旋转，因此测量可能很困难，我们的分析中也包括腋窝囊周水肿。包括腋窝隐窝发现的特异性增加表明，腋窝隐匿受累可被视为滑膜炎和旋转肌间期纤维化。常规非对比MRI肩部检查提供多个成像平面和脂肪抑制和非脂肪抑制序列，是评估粘连性囊炎的理想算法方法。因此，使用特定的MRI发现和MRI发现的星座是可行的，以最大限度地准确诊断粘连性囊炎。

根据我们的研究结果，粘着性囊炎的临床标准可与实际的1、2或3标准非对比MRI系统结合使用，以临床怀疑水平为指导。在临床高度怀疑粘着性关节囊炎的情况下，单独发现喙肱韧带增厚>2mm对确诊具有很强的敏感性。相反，喙肱韧带厚度等于或小于2mm有助于排除粘连性囊炎。在粘连性囊炎的中度临床怀疑中，增加旋转肌间隔脂肪浸润可能有助于提高诊断的可信度。在粘着性囊炎的非典型或低初始临床怀疑的情况下，腋窝隐窝增厚/水肿以及肩袖间隙发现的喙肱韧带增厚和肩袖间隙脂肪浸润的一系列发现强烈提示了对粘连性囊炎的诊断具有高度特异性，并可能引起对粘连性囊炎的怀疑。在我们的研究所连续500次常规MRI肩部检查中实施该系统，我们的MRI报告中的粘连性关节囊炎患病率为6.8%，与骨科肩部诊所中粘连性关节炎的估计患病率5–6%相似[8,19]。我们的研究有几个局限性。在我们的研究中没有关节镜或手术病理相关性。然而，由于粘连性囊炎通常不通过手术治疗，因此使用了基于亚专业体检的临床粘连性囊病参考标准，这在最近的骨科文献中已被接受。鉴于目前对粘连性囊炎的治疗主要是针对物理治疗和经皮类固醇/麻醉剂注射，组织学样本无法用于分析。此外，鉴于我们研究的回顾性设计，成像时关于粘连性囊炎临床阶段的信息不可用于审查。

5.结论

总之，结合适当的临床标准，常规非对比肩MRI可以并且应该准确一致地诊断粘连性囊炎。喙肱韧带增厚的发现显示出对粘连性囊炎的高度敏感性，而喙肱骨韧带增厚、旋转肌间隙浸润的甲状腺炎下脂肪和腋窝隐窝增厚/水肿对粘连性囊炎具有高度特异性。